Despite the explosive growth of AI and the technologies built upon it, predicting and inferring the sub-optimal behavior of users or human collaborators remains a critical challenge. In many cases, such behaviors are not a result of irrationality, but rather a rational decision made given inherent cognitive bounds and biased beliefs about the world. In this paper, we formally introduce a class of computational-rational (CR) user models for cognitively-bounded agents acting optimally under biased beliefs. The key novelty lies in explicitly modeling how a bounded memory process leads to a dynamically inconsistent and biased belief state and, consequently, sub-optimal sequential decision-making. We address the challenge of identifying the latent user-specific bound and inferring biased belief states from passive observations on the fly. We argue that for our formalized CR model family with an explicit and parameterized cognitive process, this challenge is tractable. To support our claim, we propose an efficient online inference method based on nested particle filtering that simultaneously tracks the user's latent belief state and estimates the unknown cognitive bound from a stream of observed actions. We validate our approach in a representative navigation task using memory decay as an example of a cognitive bound. With simulations, we show that (1) our CR model generates intuitively plausible behaviors corresponding to different levels of memory capacity, and (2) our inference method accurately and efficiently recovers the ground-truth cognitive bounds from limited observations (less than 100 steps). We further demonstrate how this approach provides a principled foundation for developing adaptive AI assistants, enabling adaptive assistance that accounts for the user's memory limitations.

Publisher Copyright: © 2026 The Author(s) | openaire: EC/H2020/951847/EU//ELISE

Simulation-based inference (SBI) is rapidly becoming the preferred framework for estimating parameters of intractable models in science and engineering. A significant challenge in this context is the large computational cost of simulating data from complex models, and the fact that this cost often depends on parameter values. We therefore propose \emph{cost-aware SBI methods} which can significantly reduce the cost of existing sampling-based SBI methods, such as neural SBI and approximate Bayesian computation. This is achieved through a combination of rejection and self-normalised importance sampling, which significantly reduces the number of expensive simulations needed. Our approach is studied extensively on models from epidemiology to telecommunications engineering, where we obtain significant reductions in the overall cost of inference.

Amortized meta-learning methods based on pre-training have propelled fields like natural language processing and vision. Transformer-based neural processes and their variants are leading models for probabilistic meta-learning with a tractable objective. Often trained on synthetic data, these models implicitly capture essential latent information in the data-generation process. However, existing methods do not allow users to flexibly inject (condition on) and extract (predict) this probabilistic latent information at runtime, which is key to many tasks. We introduce the Amortized Conditioning Engine (ACE), a new transformer-based meta-learning model that explicitly represents latent variables of interest. ACE affords conditioning on both observed data and interpretable latent variables, the inclusion of priors at runtime, and outputs predictive distributions for discrete and continuous data and latents. We show ACE’s practical utility across diverse tasks such as image completion and classification, Bayesian optimization, and simulation-based inference, demonstrating how a general conditioning framework can replace task-specific solutions.

Several generative models with elaborate training and sampling procedures have been proposed to accelerate structure-based drug design (SBDD); however, their empirical performance turns out to be suboptimal. We seek to better understand this phenomenon from both theoretical and empirical perspectives. Since most of these models apply graph neural networks (GNNs), one may suspect that they inherit the representational limitations of GNNs. We analyze this aspect, establishing the first such results for protein-ligand complexes. A plausible counterview may attribute the underperformance of these models to their excessive parameterizations, inducing expressivity at the expense of generalization. We investigate this possibility with a simple metric-aware approach that learns an economical surrogate for affinity to infer an unlabelled molecular graph and optimizes for labels conditioned on this graph and molecular properties. The resulting model achieves state-of-the-art results using 100x fewer trainable parameters and affords up to 1000x speedup. Collectively, our findings underscore the need to reassess and redirect the existing paradigm and efforts for SBDD. Code is available at \url{https://github.com/rafalkarczewski/SimpleSBDD.}

In the last decade, we have witnessed the introduction of several novel deep neural network (DNN) architectures exhibiting ever-increasing performance across diverse tasks. Explaining the upward trend of their performance, however, remains difficult as different DNN architectures of comparable depth and width – common factors associated with their expressive power – may exhibit a drastically different performance even when trained on the same dataset. In this paper, we introduce the concept of the non-linearity signature of DNN, the first theoretically sound solution for approximately measuring the non-linearity of deep neural networks. Built upon a score derived from closed-form optimal transport mappings, this signature provides a better understanding of the inner workings of a wide range of DNN architectures and learning paradigms, with a particular emphasis on the computer vision task. We provide extensive experimental results that highlight the practical usefulness of the proposed non-linearity signature and its potential for long-reaching implications. The code for our work is available at https://github.com/qbouniot/AffScoreDeep.

Generative Flow Networks (GFlowNets) are an emerging class of sampling methods for distributions over discrete and compositional objects, e.g., graphs. In spite of their remarkable success in problems such as drug discovery and phylogenetic inference, the question of when and whether GFlowNets learn to sample from the target distribution remains underexplored. To tackle this issue, we first assess the extent to which a violation of the detailed balance of the underlying flow network might hamper the correctness of GFlowNet's sampling distribution. In particular, we demonstrate that the impact of an imbalanced edge on the model's accuracy is influenced by the total amount of flow passing through it and, as a consequence, is unevenly distributed across the network. We also argue that, depending on the parameterization, imbalance may be inevitable. In this regard, we consider the problem of sampling from distributions over graphs with GFlowNets parameterized by graph neural networks (GNNs) and show that the representation limits of GNNs delineate which distributions these GFlowNets can approximate. Lastly, we address these limitations by proposing a theoretically sound and computationally tractable metric for assessing GFlowNets, experimentally showing it is a better proxy for correctness than popular evaluation protocols.

Conventional wisdom attributes the success of Generative Flow Networks (GFlowNets) to their ability to exploit the compositional structure of the sample space for learning generalizable flow functions (Bengio et al., 2021). Despite the abundance of empirical evidence, formalizing this belief with verifiable non-vacuous statistical guarantees has remained elusive. We address this issue with the first data-dependent generalization bounds for GFlowNets. We also elucidate the negative impact of the state space size on the generalization performance of these models via Azuma-Hoeffding-type oracle PAC-Bayesian inequalities. We leverage our theoretical insights to design a novel distributed learning algorithm for GFlowNets, which we call *Subgraph Asynchronous Learning* (SAL). In a nutshell, SAL utilizes a divide-and-conquer strategy: multiple GFlowNets are trained in parallel on smaller subnetworks of the flow network, and then aggregated with an additional GFlowNet that allocates appropriate flow to each subnetwork. Our experiments with synthetic and real-world problems demonstrate the benefits of SAL over centralized training in terms of mode coverage and distribution matching.

Preferential Bayesian Optimization (PBO) is a sample-efficient method to learn latent user utilities from preferential feedback over a pair of designs. It relies on a statistical surrogate model for the latent function, usually a Gaussian process, and an acquisition strategy to select the next candidate pair to get user feedback on. Due to the non-conjugacy of the associated likelihood, every PBO step requires a significant amount of computations with various approximate inference techniques. This computational overhead is incompatible with the way humans interact with computers, hindering the use of PBO in real-world cases. Building on the recent advances of amortized BO, we propose to circumvent this issue by fully amortizing PBO, meta-learning both the surrogate and the acquisition function. Our method comprises a novel transformer neural process architecture, trained using reinforcement learning and tailored auxiliary losses. On a benchmark composed of synthetic and real-world datasets, our method is several orders of magnitude faster than the usual Gaussian process-based strategies and often outperforms them in accuracy.

An inherent fragility of quadrotor systems stems from model inaccuracies and external disturbances. These factors hinder performance and compromise the stability of the system, making precise control challenging. Existing model-based approaches either make deterministic assumptions, utilize Gaussian-based representations of uncertainty, or rely on nominal models, all of which often fall short in capturing the complex, multimodal nature of real-world dynamics. This work introduces DroneDiffusion, a novel framework that leverages conditional diffusion models to learn quadrotor dynamics, formulated as a sequence generation task. DroneDiffusion achieves superior generalization to unseen, complex scenarios by capturing the temporal nature of uncertainties and mitigating error propagation. We integrate the learned dynamics with an adaptive controller for trajectory tracking with stability guarantees. Extensive experiments in both simulation and real-world flights demonstrate the robustness of the framework across a range of scenarios, including unfamiliar flight paths and varying payloads, velocities, and wind disturbances. Project page: https://sites.google.com/view/dronediffusion.

Many critical applications, from autonomous scientific discovery to personalized medicine, demand systems that can both strategically acquire the most informative data and instantaneously perform inference based upon it. While amortized methods for Bayesian inference and experimental design offer part of the solution, neither approach is optimal in the most general and challenging task, where new data needs to be collected for instant inference. To tackle this issue, we introduce the Amortized Active Learning and Inference Engine (ALINE), a unified framework for amortized Bayesian inference and active data acquisition. ALINE leverages a transformer architecture trained via reinforcement learning with a reward based on self-estimated information gain provided by its own integrated inference component. This allows it to strategically query informative data points while simultaneously refining its predictions. Moreover, ALINE can selectively direct its querying strategy towards specific subsets of model parameters or designated predictive tasks, optimizing for posterior estimation, data prediction, or a mixture thereof. Empirical results on regression-based active learning, classical Bayesian experimental design benchmarks, and a psychometric model with selectively targeted parameters demonstrate that ALINE delivers both instant and accurate inference along with efficient selection of informative points.

Steering cooperative multi-agent reinforcement learning (MARL) towards desired outcomes is challenging, particularly when the global guidance from a human on the whole multi-agent system is impractical in a large-scale MARL. On the other hand, designing external mechanisms (e.g., intrinsic rewards and human feedback) to coordinate agents mostly relies on empirical studies, lacking a easy-to-use research tool. In this work, we employ multi-agent influence diagrams (MAIDs) as a graphical framework to address the above issues. First, we introduce the concept of MARL interaction paradigms (orthogonal to MARL learning paradigms), using MAIDs to analyze and visualize both unguided self-organization and global guidance mechanisms in MARL. Then, we design a new MARL interaction paradigm, referred to as the targeted intervention paradigm that is applied to only a single targeted agent, so the problem of global guidance can be mitigated. In implementation, we introduce a causal inference technique—referred to as Pre-Strategy Intervention (PSI)—to realize the targeted intervention paradigm. Since MAIDs can be regarded as a special class of causal diagrams, a composite desired outcome that integrates the primary task goal and an additional desired outcome can be achieved by maximizing the corresponding causal effect through the PSI. Moreover, the bundled relevance graph analysis of MAIDs provides a tool to identify whether an MARL learning paradigm is workable under the design of an MARL interaction paradigm. In experiments, we demonstrate the effectiveness of our proposed targeted intervention, and verify the result of relevance graph analysis.

Gaussian processes (GPs) are widely used for regression and optimization tasks such as Bayesian optimization (BO) due to their expressiveness and principled uncertainty estimates. However, in settings with large datasets corrupted by outliers, standard GPs and their sparse approximations struggle with computational tractability and robustness. We introduce Robust Computation-aware Gaussian Process (RCaGP), a novel GP model that jointly addresses these challenges by combining a principled treatment of approximation-induced uncertainty with robust generalized Bayesian updating. The key insight is that robustness and approximation-awareness are not orthogonal but intertwined: approximations can exacerbate the impact of outliers, and mitigating one without the other is insufficient. Unlike previous work that focuses narrowly on either robustness or approximation quality, RCaGP combines both in a principled and scalable framework, thus effectively managing both outliers and computational uncertainties introduced by approximations such as low-rank matrix multiplications. Our model ensures more conservative and reliable uncertainty estimates, a property we rigorously demonstrate. Additionally, we establish a robustness property and show that the mean function is key to preserving it, motivating a tailored model selection scheme for robust mean functions. Empirical results confirm that solving these challenges jointly leads to superior performance across both clean and outlier-contaminated settings, both on regression and high-throughput Bayesian optimization benchmarks.

Generalization outside the scope of one’s training data requires leveraging prior knowledge about the effects that transfer, and the effects that don’t, between different data sources. Transfer learning is a framework for specifying and refining this knowledge about sets of source (training) and target (prediction) data. A challenging open problem is addressing the empirical phenomenon of negative transfer, whereby the transfer learner performs worse on the target data after taking the source data into account than before. We first introduce a Bayesian perspective on negative transfer, and then a method to address it. The key insight from our formulation is that negative transfer can stem from misspecified prior information about non-transferable causes of the source data. Our proposed method, proxy-informed robust method for probabilistic transfer learning (PROMPT), does not require prior knowledge of the source data (the data sources may be "unknown"). PROMPT is thus applicable when differences between tasks are unobserved, such as in the presence of latent confounders. Moreover, the learner need not have access to observations in the target task (may not have the ability to "fine-tune"), and instead makes use of proxy (indirect) information. Our theoretical results show that the threat of negative transfer does not depend on the informativeness of the proxy information, highlighting the usefulness of PROMPT in cases where only noisy indirect information, such as human feedback, is available.

Uncertainty-aware user modeling is crucial for designing AI systems that adapt to users in real-time while addressing privacy concerns. This paper proposes a novel framework for privacy-preserving probabilistic user modeling that integrates uncertainty quantification and differential privacy (DP). Building on neural processes (NPs), a scalable latent variable probabilistic model, we enable meta-learning for user behaviour prediction under privacy constraints. By employing differentially private stochastic gradient descent (DP-SGD), our method achieves rigorous privacy guarantees while preserving predictive accuracy. Unlike prior work, which primarily addresses privacy-preserving learning for convex or smooth functions, we establish theoretical guarantees for non-convex objectives, focusing on the utility-privacy trade-offs inherent in uncertainty-aware models. Through extensive experiments, we demonstrate that our approach achieves competitive accuracy under stringent privacy budgets. Our results showcase the potential of privacy-preserving probabilistic user models to enable trustworthy AI systems in real-world interactive applications.

Randomized controlled trials (RCTs) are the accepted standard for treatment effect estimation but they can be infeasible due to ethical reasons and prohibitive costs. Single-arm trials, where all patients belong to the treatment group, can be a viable alternative but require access to an external control group. We propose an identifiable deep latent-variable model for this scenario that can also account for missing covariate observations by modeling their structured missingness patterns. Our method uses amortized variational inference to learn both group-specific and identifiable shared latent representations, which can subsequently be used for {\em (i)} patient matching if treatment outcomes are not available for the treatment group, or for {\em (ii)} direct treatment effect estimation assuming outcomes are available for both groups. We evaluate the model on a public benchmark as well as on a data set consisting of a published RCT study and real-world electronic health records. Compared to previous methods, our results show improved performance both for direct treatment effect estimation as well as for effect estimation via patient matching.

Deep Ensembles (DEs) demonstrate improved accuracy, calibration and robustness to perturbations over single neural networks partly due to their functional diversity. Particle-based variational inference (ParVI) methods enhance diversity by formalizing a repulsion term based on a network similarity kernel. However, weight-space repulsion is inefficient due to over-parameterization, while direct function-space repulsion has been found to produce little improvement over DEs. To sidestep these difficulties, we propose First-order Repulsive Deep Ensemble (FoRDE), an ensemble learning method based on ParVI, which performs repulsion in the space of first-order input gradients. As input gradients uniquely characterize a function up to translation and are much smaller in dimension than the weights, this method guarantees that ensemble members are functionally different. Intuitively, diversifying the input gradients encourages each network to learn different features, which is expected to improve the robustness of an ensemble. Experiments on image classification datasets and transfer learning tasks show that FoRDE significantly outperforms the gold-standard DEs and other ensemble methods in accuracy and calibration under covariate shift due to input perturbations.

GFlowNets are a promising alternative to MCMC sampling for discrete compositional random variables. Training GFlowNets requires repeated evaluations of the unnormalized target distribution, or reward function. However, for large-scale posterior sampling, this may be prohibitive since it incurs traversing the data several times. Moreover, if the data are distributed across clients, employing standard GFlowNets leads to intensive client-server communication. To alleviate both these issues, we propose embarrassingly parallel GFlowNet (EP-GFlowNet). EP-GFlowNet is a provably correct divide-and-conquer method to sample from product distributions of the form $R(\cdot) \propto R_1(\cdot) ... R_N(\cdot)$ — e.g., in parallel or federated Bayes, where each $R_n$ is a local posterior defined on a data partition. First, in parallel, we train a local GFlowNet targeting each $R_n$ and send the resulting models to the server. Then, the server learns a global GFlowNet by enforcing our newly proposed aggregating balance condition, requiring a single communication step. Importantly, EP-GFlowNets can also be applied to multi-objective optimization and model reuse. Our experiments illustrate the effectiveness of EP-GFlowNets on multiple tasks, including parallel Bayesian phylogenetics, multi-objective multiset and sequence generation, and federated Bayesian structure learning.

Ad hoc teamwork poses a challenging problem, requiring the design of an agent to collaborate with teammates without prior coordination or joint training. Open ad hoc teamwork (OAHT) further complicates this challenge by considering environments with a changing number of teammates, referred to as open teams. One promising solution in practice to this problem is leveraging the generalizability of graph neural networks to handle an unrestricted number of agents with various agent-types, named graph-based policy learning (GPL). However, its joint Q-value representation over a coordination graph lacks convincing explanations. In this paper, we establish a new theory to understand the representation of the joint Q-value for OAHT and its learning paradigm, through the lens of cooperative game theory. Building on our theory, we propose a novel algorithm named CIAO, based on GPL's framework, with additional provable implementation tricks that can facilitate learning. The demos of experimental results are available on https://sites.google.com/view/ciao2024, and the code of experiments is published on https://github.com/hsvgbkhgbv/CIAO.

Many critical decisions, such as personalized medical diagnoses and product pricing, are made based on insights gained from designing, observing, and analyzing a series of experiments. This highlights the crucial role of experimental design, which goes beyond merely collecting information on system parameters as in traditional Bayesian experimental design (BED), but also plays a key part in facilitating downstream decision-making. Most recent BED methods use an amortized policy network to rapidly design experiments. However, the information gathered through these methods is suboptimal for down-the-line decision-making, as the experiments are not inherently designed with downstream objectives in mind. In this paper, we present an amortized decision-aware BED framework that prioritizes maximizing downstream decision utility. We introduce a novel architecture, the Transformer Neural Decision Process (TNDP), capable of instantly proposing the next experimental design, whilst inferring the downstream decision, thus effectively amortizing both tasks within a unified workflow. We demonstrate the performance of our method across several tasks, showing that it can deliver informative designs and facilitate accurate decision-making.

Time series data are essential in a wide range of machine learning (ML) applications. However, temporal data are often scarce or highly sensitive, limiting data sharing and the use of data-intensive ML methods. A possible solution to this problem is the generation of synthetic datasets that resemble real data. In this work, we introduce Time Series Generative Modeling (TSGM), an open-source framework for the generative modeling and evaluation of synthetic time series datasets. TSGM includes a broad repertoire of machine learning methods: generative models, probabilistic, simulation-based approaches, and augmentation techniques. The framework enables users to evaluate the quality of the produced data from different angles: similarity, downstream effectiveness, predictive consistency, diversity, fairness, and privacy. TSGM is extensible and user-friendly, which allows researchers to rapidly implement their own methods and compare them in a shareable environment. The framework has been tested on open datasets and in production and proved to be beneficial in both cases. https://github.com/AlexanderVNikitin/tsgm

Preference learning methods make use of models of human choice in order to infer the latent utilities that underlie human behavior. However, accurate modeling of human choice behavior is challenging due to a range of context effects that arise from how humans contrast and evaluate options. Cognitive science has proposed several models that capture these intricacies but, due to their intractable nature, work on preference learning has, in practice, had to rely on tractable but simplified variants of the well-known Bradley-Terry model. In this paper, we take one state-of-the-art intractable cognitive model and propose a tractable surrogate that is suitable for deployment in preference learning. We then introduce a mechanism for fitting the surrogate to human data and extend it to account for data that cannot be explained by the original cognitive model. We demonstrate on large-scale human data that this model produces significantly better inferences on static and actively elicited data than existing Bradley-Terry variants. We further show in simulation that when using this model for preference learning, we can significantly improve utility in a range of real-world tasks.

Deep neural networks (DNNs) excel on clean images but struggle with corrupted ones. Incorporating specific corruptions into the data augmentation pipeline can improve robustness to those corruptions but may harm performance on clean images and other types of distortion. In this paper, we introduce an alternative approach that improves the robustness of DNNs to a wide range of corruptions without compromising accuracy on clean images. We first demonstrate that input perturbations can be mimicked by multiplicative perturbations in the weight space. Leveraging this, we propose Data Augmentation via Multiplicative Perturbation (DAMP), a training method that optimizes DNNs under random multiplicative weight perturbations. We also examine the recently proposed Adaptive Sharpness-Aware Minimization (ASAM) and show that it optimizes DNNs under adversarial multiplicative weight perturbations. Experiments on image classification datasets (CIFAR-10/100, TinyImageNet and ImageNet) and neural network architectures (ResNet50, ViT-S/16, ViT-B/16) show that DAMP enhances model generalization performance in the presence of corruptions across different settings. Notably, DAMP is able to train a ViT-S/16 on ImageNet from scratch, reaching the top-1 error of 23.7% which is comparable to ResNet50 without extensive data augmentations.

Contextual Bayesian Optimization (CBO) efficiently optimizes black-box functions with respect to design variables, while simultaneously integrating _contextual_ information regarding the environment, such as experimental conditions. However, the relevance of contextual variables is not necessarily known beforehand. Moreover, contextual variables can sometimes be optimized themselves at additional cost, a setting overlooked by current CBO algorithms. Cost-sensitive CBO would simply include optimizable contextual variables as part of the design variables based on their cost. Instead, we adaptively select a subset of contextual variables to include in the optimization, based on the trade-off between their _relevance_ and the additional cost incurred by optimizing them compared to leaving them to be determined by the environment. We learn the relevance of contextual variables by sensitivity analysis of the posterior surrogate model while minimizing the cost of optimization by leveraging recent developments on early stopping for BO. We empirically evaluate our proposed Sensitivity-Analysis-Driven Contextual BO (_SADCBO_) method against alternatives on both synthetic and real-world experiments, together with extensive ablation studies, and demonstrate a consistent improvement across examples.

In many settings, such as scientific inference, optimization, and transfer learning, the learner has a well-defined objective, which can be treated as estimation of a target parameter, and no intrinsic interest in characterizing the entire data-generating process. Usually, the learner must also contend with additional sources of uncertainty or variables — with nuisance parameters. Bayesian active learning, or sequential optimal experimental design, can straightforwardly accommodate the presence of nuisance parameters, and so is a natural active learning framework for such problems. However, the introduction of nuisance parameters can lead to bias in the Bayesian learner’s estimate of the target parameters, a phenomenon we refer to as negative interference. We characterize the threat of negative interference and how it fundamentally changes the nature of the Bayesian active learner’s task. We show that the extent of negative interference can be extremely large, and that accurate estimation of the nuisance parameters is critical to reducing it. The Bayesian active learner is confronted with a dilemma: whether to spend a finite acquisition budget in pursuit of estimation of the target or of the nuisance parameters. Our setting encompasses Bayesian transfer learning as a special case, and our results shed light on the phenomenon of negative transfer between learning environments.

Our digital life consists of activities that are organized around tasks and exhibit different user states in the digital contexts around these activities. Previous works have shown that digital activity monitoring can be used to predict entities that users will need to perform digital tasks. There have been methods developed to automatically detect the tasks of a user. However, these studies typically support only specific applications and tasks, and relatively little research has been conducted on real-life digital activities. This article introduces user state modeling and prediction with contextual information captured as entities, recorded from real-world digital user behavior, called entity footprinting —a system that records users’ digital activities on their screens and proactively provides useful entities across application boundaries without requiring explicit query formulation. Our methodology is to detect contextual user states using latent representations of entities occurring in digital activities. Using topic models and recurrent neural networks, the model learns the latent representation of concurrent entities and their sequential relationships. We report a field study in which the digital activities of 13 people were recorded continuously for 14 days. The model learned from this data is used to (1) predict contextual user states and (2) predict relevant entities for the detected states. The results show improved user state detection accuracy and entity prediction performance compared to static, heuristic, and basic topic models. Our findings have implications for the design of proactive recommendation systems that can implicitly infer users’ contextual state by monitoring users’ digital activities and proactively recommending the right information at the right time.

In sequential machine teaching, a teacher’s objective is to provide the optimal sequence of inputs to sequential learners in order to guide them towards the best model. However, this teaching objective considers a restricted class of learners with fixed inductive biases. In this paper, we extend the machine teaching framework to learners that can improve their inductive biases, represented as latent internal states, in order to generalize to new datasets. We introduce a novel framework in which learners’ inductive biases may change with the teaching interaction, which affects the learning performance in future tasks. In order to teach such learners, we propose a multi-objective control approach that takes the future performance of the learner after teaching into account. This framework provides tools for modelling learners with internal states, humans and meta-learning algorithms alike. Furthermore, we distinguish manipulative teaching, which can be done by effectively hiding data and also used for indoctrination, from teaching to learn which aims to help the learner become better at learning from new datasets in the absence of a teacher. Our empirical results demonstrate that our framework is able to reduce the number of required tasks for online meta-learning, and increases independent learning performance of simulated human users in future tasks.

We consider the problem of creating assistants that can help agents solve new sequential decision problems, assuming the agent is not able to specify the reward function explicitly to the assistant. Instead of acting in place of the agent as in current automation-based approaches, we give the assistant an advisory role and keep the agent in the loop as the main decision maker. The difficulty is that we must account for potential biases of the agent which may cause it to seemingly irrationally reject advice. To do this we introduce a novel formalization of assistance that models these biases, allowing the assistant to infer and adapt to them. We then introduce a new method for planning the assistant's actions which can scale to large decision making problems. We show experimentally that our approach adapts to these agent biases, and results in higher cumulative reward for the agent than automation-based alternatives. Lastly, we show that an approach combining advice and automation outperforms advice alone at the cost of losing some safety guarantees.

Bayesian optimization (BO) is a powerful framework for optimizing black-box, expensive-to-evaluate functions. Over the past decade, many algorithms have been proposed to integrate cheaper, lower-fidelity approximations of the objective function into the optimization process, with the goal of converging towards the global optimum at a reduced cost. This task is generally referred to as multi-fidelity Bayesian optimization (MFBO). However, MFBO algorithms can lead to higher optimization costs than their vanilla BO counterparts, especially when the low-fidelity sources are poor approximations of the objective function, therefore defeating their purpose. To address this issue, we propose rMFBO (robust MFBO), a methodology to make any GP-based MFBO scheme robust to the addition of unreliable information sources. rMFBO comes with a theoretical guarantee that its performance can be bound to its vanilla BO analog, with high controllable probability. We demonstrate the effectiveness of the proposed methodology on a number of numerical benchmarks, outperforming earlier MFBO methods on unreliable sources. We expect rMFBO to be particularly useful to reliably include human experts with varying knowledge within BO processes.

While generation of synthetic data under differential privacy (DP) has received a lot of attention in the data privacy community, analysis of synthetic data has received much less. Existing work has shown that simply analysing DP synthetic data as if it were real does not produce valid inferences of population-level quantities. For example, confidence intervals become too narrow, which we demonstrate with a simple experiment. We tackle this problem by combining synthetic data analysis techniques from the field of multiple imputation (MI), and synthetic data generation using noise-aware (NA) Bayesian modeling into a pipeline NA+MI that allows computing accurate uncertainty estimates for population-level quantities from DP synthetic data. To implement NA+MI for discrete data generation using the values of marginal queries, we develop a novel noise-aware synthetic data generation algorithm NAPSU-MQ using the principle of maximum entropy. Our experiments demonstrate that the pipeline is able to produce accurate confidence intervals from DP synthetic data. The intervals become wider with tighter privacy to accurately capture the additional uncertainty stemming from DP noise.

Likelihood-free inference methods typically make use of a distance between simulated and real data. A common example is the maximum mean discrepancy (MMD), which has previously been used for approximate Bayesian computation, minimum distance estimation, generalised Bayesian inference, and within the nonparametric learning framework. The MMD is commonly estimated at a root-$m$ rate, where $m$ is the number of simulated samples. This can lead to significant computational challenges since a large $m$ is required to obtain an accurate estimate, which is crucial for parameter estimation. In this paper, we propose a novel estimator for the MMD with significantly improved sample complexity. The estimator is particularly well suited for computationally expensive smooth simulators with low- to mid-dimensional inputs. This claim is supported through both theoretical results and an extensive simulation study on benchmark simulators.

Learning policies from multiple demonstrators is often difficult because different individuals perform the same task differently due to hidden factors such as preferences. In the context of policy learning, this leads to multimodal policies. Existing policy learning methods often converge to a single solution mode, failing to capture the diversity in the solution space. In this paper, we introduce an imitation-guided reinforcement learning framework to solve the multimodal policy learning problem from a limited number of state-only demonstrations. Then, we propose LfBD (Learning from Behaviourally diverse Demonstration), an algorithm that builds a parameterised solution space to capture the variability in the behaviour space defined by demonstrations. To this end, we define a projection function based on the state density distributions from demonstrations to define such space. Our goal is not only to learn how to solve the task as the human demonstrator but also to extrapolate beyond the provided demonstrations. In addition, we show that with our method, we can perform a post-hoc policy search in the built solution space to recover policies that satisfy specific constraints or to find a policy that matches a given (state-only) behaviour.

High training costs of generative models and the need to fine-tune them for specific tasks have created a strong interest in model reuse and composition. A key challenge in composing iterative generative processes, such as GFlowNets and diffusion models, is that to realize the desired target distribution, all steps of the generative process need to be coordinated, and satisfy delicate balance conditions. In this work, we propose Compositional Sculpting: a general approach for defining compositions of iterative generative processes. We then introduce a method for sampling from these compositions built on classifier guidance. We showcase ways to accomplish compositional sculpting in both GFlowNets and diffusion models. We highlight two binary operations $\\unicode{x2014}$ the $\\textit{harmonic mean}\\unicode{x00A0}(p_1 \\otimes p_2$) and the $\\textit{contrast}\\unicode{x00A0}(p_1 \\,\\unicode{x25D1}\\,\\, p_2$) between pairs, and the generalization of these operations to multiple component distributions. We offer empirical results on image and molecular generation tasks. Project codebase: https://github.com/timgaripov/compositional-sculpting.

Simulation-based inference (SBI) methods such as approximate Bayesian computation (ABC), synthetic likelihood, and neural posterior estimation (NPE) rely on simulating statistics to infer parameters of intractable likelihood models. However, such methods are known to yield untrustworthy and misleading inference outcomes under model misspecification, thus hindering their widespread applicability. In this work, we propose the first general approach to handle model misspecification that works across different classes of SBI methods. Leveraging the fact that the choice of statistics determines the degree of misspecification in SBI, we introduce a regularized loss function that penalizes those statistics that increase the mismatch between the data and the model. Taking NPE and ABC as use cases, we demonstrate the superior performance of our method on high-dimensional time-series models that are artificially misspecified. We also apply our method to real data from the field of radio propagation where the model is known to be misspecified. We show empirically that the method yields robust inference in misspecified scenarios, whilst still being accurate when the model is well-specified.

Conditional Neural Processes (CNPs) are a class of metalearning models popular for combining the runtime efficiency of amortized inference with reliable uncertainty quantification. Many relevant machine learning tasks, such as in spatio-temporal modeling, Bayesian Optimization and continuous control, inherently contain equivariances – for example to translation – which the model can exploit for maximal performance. However, prior attempts to include equivariances in CNPs do not scale effectively beyond two input dimensions. In this work, we propose Relational Conditional Neural Processes (RCNPs), an effective approach to incorporate equivariances into any neural process model. Our proposed method extends the applicability and impact of equivariant neural processes to higher dimensions. We empirically demonstrate the competitive performance of RCNPs on a large array of tasks naturally containing equivariances.

Probabilistic user modeling is essential for building machine learning systems in the ubiquitous cases with humans in the loop. However, modern advanced user models, often designed as cognitive behavior simulators, are incompatible with modern machine learning pipelines and computationally prohibitive for most practical applications. We address this problem by introducing widely-applicable differentiable surrogates for bypassing this computational bottleneck; the surrogates enable computationally efficient inference with modern cognitive models. We show experimentally that modeling capabilities comparable to the only available solution, existing likelihood-free inference methods, are achievable with a computational cost suitable for online applications. Finally, we demonstrate how AI-assistants can now use cognitive models for online interaction in a menu-search task, which has so far required hours of computation during interaction.

MOTIVATION: Existing methods for simulating synthetic genotype and phenotype datasets have limited scalability, constraining their usability for large-scale analyses. Moreover, a systematic approach for evaluating synthetic data quality and a benchmark synthetic dataset for developing and evaluating methods for polygenic risk scores are lacking. RESULTS: We present HAPNEST, a novel approach for efficiently generating diverse individual-level genotypic and phenotypic data. In comparison to alternative methods, HAPNEST shows faster computational speed and a lower degree of relatedness with reference panels, while generating datasets that preserve key statistical properties of real data. These desirable synthetic data properties enabled us to generate 6.8 million common variants and nine phenotypes with varying degrees of heritability and polygenicity across 1 million individuals. We demonstrate how HAPNEST can facilitate biobank-scale analyses through the comparison of seven methods to generate polygenic risk scoring across multiple ancestry groups and different genetic architectures. AVAILABILITY AND IMPLEMENTATION: A synthetic dataset of 1 008 000 individuals and nine traits for 6.8 million common variants is available at https://www.ebi.ac.uk/biostudies/studies/S-BSST936. The HAPNEST software for generating synthetic datasets is available as Docker/Singularity containers and open source Julia and C code at https://github.com/intervene-EU-H2020/synthetic_data.

Gaussian processes (GPs) provide a principled and direct approach for inference and learning on graphs. However, the lack of justified graph kernels for spatio-temporal modelling has held back their use in graph problems. We leverage an explicit link between stochastic partial differential equations (SPDEs) and GPs on graphs, introduce a framework for deriving graph kernels via SPDEs, and derive non-separable spatio-temporal graph kernels that capture interaction across space and time. We formulate the graph kernels for the stochastic heat equation and wave equation. We show that by providing novel tools for spatio-temporal GP modelling on graphs, we outperform pre-existing graph kernels in real-world applications that feature diffusion, oscillation, and other complicated interactions.

Embarrassingly parallel Markov Chain Monte Carlo (MCMC) exploits parallel computing to scale Bayesian inference to large datasets by using a two-step approach. First, MCMC is run in parallel on (sub)posteriors defined on data partitions. Then, a server combines local results. While efficient, this framework is very sensitive to the quality of subposterior sampling. Common sampling problems such as missing modes or misrepresentation of low-density regions are amplified – instead of being corrected – in the combination phase, leading to catastrophic failures. In this work, we propose a novel combination strategy to mitigate this issue. Our strategy, Parallel Active Inference (PAI), leverages Gaussian Process (GP) surrogate modeling and active learning. After fitting GPs to subposteriors, PAI (i) shares information between GP surrogates to cover missing modes; and (ii) uses active sampling to individually refine subposterior approximations. We validate PAI in challenging benchmarks, including heavy-tailed and multi-modal posteriors and a real-world application to computational neuroscience. Empirical results show that PAI succeeds where previous methods catastrophically fail, with a small communication overhead.

Approximate Bayesian computation (ABC) is a popular likelihood-free inference method for models with intractable likelihood functions. As ABC methods usually rely on comparing summary statistics of observed and simulated data, the choice of the statistics is crucial. This choice involves a trade-off between loss of information and dimensionality reduction, and is often determined based on domain knowledge. However, handcrafting and selecting suitable statistics is a laborious task involving multiple trial-and-error steps. In this work, we introduce an active learning method for ABC statistics selection which reduces the domain expert’s work considerably. By involving the experts, we are able to handle misspecified models, unlike the existing dimension reduction methods. Moreover, empirical results show better posterior estimates than with existing methods, when the simulation budget is limited.

Bayesian neural networks (BNNs) promise improved generalization under covariate shift by providing principled probabilistic representations of epistemic uncertainty. However, weight-based BNNs often struggle with high computational complexity of large-scale architectures and datasets. Node-based BNNs have recently been introduced as scalable alternatives, which induce epistemic uncertainty by multiplying each hidden node with latent random variables, while learning a point-estimate of the weights. In this paper, we interpret these latent noise variables as implicit representations of simple and domain-agnostic data perturbations during training, producing BNNs that perform well under covariate shift due to input corruptions. We observe that the diversity of the implicit corruptions depends on the entropy of the latent variables, and propose a straightforward approach to increase the entropy of these variables during training. We evaluate the method on out-of-distribution image classification benchmarks, and show improved uncertainty estimation of node-based BNNs under covariate shift due to input perturbations. As a side effect, the method also provides robustness against noisy training labels.

Predictive maintenance (PdM) is the task of scheduling maintenance operations based on a statistical analysis of the system's condition. We propose a human-in-the-loop PdM approach in which a machine learning system predicts future problems in sets of workstations (computers, laptops, and servers). Our system interacts with domain experts to improve predictions and elicit their knowledge. In our approach, domain experts are included in the loop not only as providers of correct labels, as in traditional active learning, but as a source of explicit decision rule feedback. The system is automated and designed to be easily extended to novel domains, such as maintaining workstations of several organizations. In addition, we develop a simulator for reproducible experiments in a controlled environment and deploy the system in a large-scale case of real-life workstations PdM with thousands of workstations for dozens of companies.

Similarity metrics such as representational similarity analysis (RSA) and centered kernel alignment (CKA) have been used to understand neural networks by comparing their layer-wise representations. However, these metrics are confounded by the population structure of data items in the input space, leading to inconsistent conclusions about the \emph{functional} similarity between neural networks, such as spuriously high similarity of completely random neural networks and inconsistent domain relations in transfer learning. We introduce a simple and generally applicable fix to adjust for the confounder with covariate adjustment regression, which improves the ability of CKA and RSA to reveal functional similarity and also retains the intuitive invariance properties of the original similarity measures. We show that deconfounding the similarity metrics increases the resolution of detecting functionally similar neural networks across domains. Moreover, in real-world applications, deconfounding improves the consistency between CKA and domain similarity in transfer learning, and increases the correlation between CKA and model out-of-distribution accuracy similarity.

Temporal graph networks (TGNs) have gained prominence as models for embedding dynamic interactions, but little is known about their theoretical underpinnings. We establish fundamental results about the representational power and limits of the two main categories of TGNs: those that aggregate temporal walks (WA-TGNs), and those that augment local message passing with recurrent memory modules (MP-TGNs). Specifically, novel constructions reveal the inadequacy of MP-TGNs and WA-TGNs, proving that neither category subsumes the other. We extend the 1-WL (Weisfeiler-Leman) test to temporal graphs, and show that the most powerful MP-TGNs should use injective updates, as in this case they become as expressive as the temporal WL. Also, we show that sufficiently deep MP-TGNs cannot benefit from memory, and MP/WA-TGNs fail to compute graph properties such as girth. These theoretical insights lead us to PINT --- a novel architecture that leverages injective temporal message passing and relative positional features. Importantly, PINT is provably more expressive than both MP-TGNs and WA-TGNs. PINT significantly outperforms existing TGNs on several real-world benchmarks.

Generating new molecules is fundamental to advancing critical applications such as drug discovery and material synthesis. Flows can generate molecules effectively by inverting the encoding process, however, existing flow models either require artifactual dequantization or specific node/edge orderings, lack desiderata such as permutation invariance, or induce discrepancy between encoding and decoding steps that necessitates post hoc validity correction. Inspired by graph PDEs, we circumvent these issues with novel continuous normalizing E(3)-equivariant flows, based on a system of coupled node ODEs, that repeatedly reconcile locally toward globally aligned densities. Our models can be cast as message passing temporal networks, and result in superlative density estimation and molecular generation. In particular, our generated samples achieve state of the art on both the standard QM9 and ZINC250K benchmarks.

Recent machine learning advances have proposed black-box estimation of \textit{unknown continuous-time system dynamics} directly from data. However, earlier works are based on approximative solutions or point estimates. We propose a novel Bayesian nonparametric model that uses Gaussian processes to infer posteriors of unknown ODE systems directly from data. We derive sparse variational inference with decoupled functional sampling to represent vector field posteriors. We also introduce a probabilistic shooting augmentation to enable efficient inference from arbitrarily long trajectories. The method demonstrates the benefit of computing vector field posteriors, with predictive uncertainty scores outperforming alternative methods on multiple ODE learning tasks.

BACKGROUND: A deep understanding of carcinogenesis at the DNA level underpins many advances in cancer prevention and treatment. Mutational signatures provide a breakthrough conceptualisation, as well as an analysis framework, that can be used to build such understanding. They capture somatic mutation patterns and at best identify their causes. Most studies in this context have focused on an inherently additive analysis, e.g. by non-negative matrix factorization, where the mutations within a cancer sample are explained by a linear combination of independent mutational signatures. However, other recent studies show that the mutational signatures exhibit non-additive interactions. RESULTS: We carefully analysed such additive model fits from the PCAWG study cataloguing mutational signatures as well as their activities across thousands of cancers. Our analysis identified systematic and non-random structure of residuals that is left unexplained by the additive model. We used hierarchical clustering to identify cancer subsets with similar residual profiles to show that both systematic mutation count overestimation and underestimation take place. We propose an extension to the additive mutational signature model-multiplicatively acting modulatory processes-and develop a maximum-likelihood framework to identify such modulatory mutational signatures. The augmented model is expressive enough to almost fully remove the observed systematic residual patterns. CONCLUSION: We suggest the modulatory processes biologically relate to sample specific DNA repair propensities with cancer or tissue type specific profiles. Overall, our results identify an interesting direction where to expand signature analysis.

Detecting predictive biomarkers from multi-omics data is important for precision medicine, to improve diagnostics of complex diseases and for better treatments. This needs substantial experimental efforts that are made difficult by the heterogeneity of cell lines and huge cost. An effective solution is to build a computational model over the diverse omics data, including genomic, molecular, and environmental information. However, choosing informative and reliable data sources from among the different types of data is a challenging problem. We propose DIVERSE, a framework of Bayesian importance-weighted tri- and bi-matrix factorization(DIVERSE3 or DIVERSE2) to predict drug responses from data of cell lines, drugs, and gene interactions. DIVERSE integrates the data sources systematically, in a step-wise manner, examining the importance of each added data set in turn. More specifically, we sequentially integrate five different data sets, which have not all been combined in earlier bioinformatic methods for predicting drug responses. Empirical experiments show that DIVERSE clearly outperformed five other methods including three state-of-the-art approaches, under cross-validation, particularly in out-of-matrix prediction, which is closer to the setting of real use cases and more challenging than simpler in-matrix prediction. Additionally, case studies for discovering new drugs further confirmed the performance advantage of DIVERSE.

In machine learning and computer vision, optimal transport has had significant success inlearning generative models and defining metric distances between structured and stochasticdata objects, that can be cast as probability measures. The key element of optimal trans-port is the so called lifting of anexactcost (distance) function, defined on the sample space,to a cost (distance) between probability measures over the sample space. However, in manyreal life applications the cost isstochastic: e.g., the unpredictable traffic flow affects the costof transportation between a factory and an outlet. To take this stochasticity into account,we introduce a Bayesian framework for inferring the optimal transport plan distributioninduced by the stochastic cost, allowing for a principled way to include prior informationand to model the induced stochasticity on the transport plans. Additionally, we tailor anHMC method to sample from the resulting transport plan posterior distribution.

Generalized linear models (GLMs) such as logistic regression are among the most widely used arms in data analyst’s repertoire and often used on sensitive datasets. A large body of prior works that investigate GLMs under differential privacy (DP) constraints provide only private point estimates of the regression coefficients, and are not able to quantify parameter uncertainty. In this work, with logistic and Poisson regression as running examples, we introduce a generic noise-aware DP Bayesian inference method for a GLM at hand, given a noisy sum of summary statistics. Quantifying uncertainty allows us to determine which of the regression coefficients are statistically significantly different from zero. We provide a previously unknown tight privacy analysis and experimentally demonstrate that the posteriors obtained from our model, while adhering to strong privacy guarantees, are close to the non-private posteriors.

Human-in-the-loop machine learning is widely used in artificial intelligence (AI) to elicit labels for data points from experts or to provide feedback on how close the predicted results are to the target. This simplifies away all the details of the decision-making process of the expert. In this work, we allow the experts to additionally produce decision rules describing their decision-making; the rules are expected to be imperfect but to give additional information. In particular, the rules can extend to new distributions, and hence enable significantly improving performance for cases where the training and testing distributions differ, such as in domain adaptation. We apply the proposed method to lifelong learning and domain adaptation problems and discuss applications in other branches of AI, such as knowledge acquisition problems in expert systems. In simulated and real-user studies, we show that decision rule elicitation improves domain adaptation of the algorithm and helps to propagate expert’s knowledge to the AI model.

Approximate Bayesian inference estimates descriptors of an intractable target distribution - in essence, an optimization problem within a family of distributions. For example, Langevin dynamics (LD) extracts asymptotically exact samples from a diffusion process because the time evolution of its marginal distributions constitutes a curve that minimizes the KL-divergence via steepest descent in the Wasserstein space. Parallel to LD, Stein variational gradient descent (SVGD) similarly minimizes the KL, albeit endowed with a novel Stein-Wasserstein distance, by deterministically transporting a set of particle samples, thus de-randomizes the stochastic diffusion process. We propose de-randomized kernel-based particle samplers to all diffusion-based samplers known as MCMC dynamics. Following previous work in interpreting MCMC dynamics, we equip the Stein-Wasserstein space with a fiber-Riemannian Poisson structure, with the capacity of characterizing a fiber-gradient Hamiltonian flow that simulates MCMC dynamics. Such dynamics discretizes into generalized SVGD (GSVGD), a Stein-type deterministic particle sampler, with particle updates coinciding with applying the diffusion Stein operator to a kernel function. We demonstrate empirically that GSVGD can de-randomize complex MCMC dynamics, which combine the advantages of auxiliary momentum variables and Riemannian structure, while maintaining the high sample quality from an interacting particle system.

Everyday digital tasks can highly benefit from systems that recommend the right information to use at the right time. However, existing solutions typically support only specific applications and tasks. In this demo, we showcase EntityBot, a system that captures context across application boundaries and recommends information entities related to the current task. The user’s digital activity is continuously monitored by capturing all content on the computer screen using optical character recognition. This includes all applications and services being used and specific to individuals’ computer usages such as instant messaging, emailing, web browsing, and word processing. A linear model is then applied to detect the user’s task context to retrieve entities such as applications, documents, contact information, and several keywords determining the task. The system has been evaluated with real-world tasks, demonstrating that the recommendation had an impact on the tasks and led to high user satisfaction.

Stochastic gradient MCMC methods, such as stochastic gradient Langevin dynamics (SGLD), employ fast but noisy gradient estimates to enable large-scale posterior sampling. Although we can easily extend SGLD to distributed settings, it suffers from two issues when applied to federated non-IID data. First, the variance of these estimates increases significantly. Second, delaying communication causes the Markov chains to diverge from the true posterior even for very simple models. To alleviate both these problems, we propose conducive gradients, a simple mechanism that combines local likelihood approximations to correct gradient updates. Notably, conducive gradients are easy to compute, and since we only calculate the approximations once, they incur negligible overhead. We apply conducive gradients to distributed stochastic gradient Langevin dynamics (DSGLD) and call the resulting method “federated stochastic gradient Langevin dynamics” (FSGLD). We demonstrate that our approach can handle delayed communication rounds, converging to the target posterior in cases where DSGLD fails. We also show that FSGLD outperforms DSGLD for non-IID federated data with experiments on metric learning and neural networks.

Drug combination therapy is a promising strategy to treat complex diseases such as cancer and infectious diseases. However, current knowledge of drug combination therapies, especially in cancer patients, is limited because of adverse drug effects, toxicity and cell line heterogeneity. Screening new drug combinations requires substantial efforts since considering all possible combinations between drugs is infeasible and expensive. Therefore, building computational approaches, particularly machine learning methods, could provide an effective strategy to overcome drug resistance and improve therapeutic efficacy. In this review, we group the state-of-the-art machine learning approaches to analyze personalized drug combination therapies into three categories and discuss each method in each category. We also present a short description of relevant databases used as a benchmark in drug combination therapies and provide a list of well-known, publicly available interactive data analysis portals. We highlight the importance of data integration on the identification of drug combinations. Finally, we address the advantages of combining multiple data sources on drug combination analysis by showing an experimental comparison.

Predicting the response of cancer cell lines to specific drugs is one of the central problems in personalized medicine, where the cell lines show diverse characteristics. Researchers have developed a variety of computational methods to discover associations between drugs and cell lines, and improved drug sensitivity analyses by integrating heterogeneous biological data. However, choosing informative data sources and methods that can incorporate multiple sources efficiently is the challenging part of successful analysis in personalized medicine. The reason is that finding decisive factors of cancer and developing methods that can overcome the problems of integrating data, such as differences in data structures and data complexities, are difficult. In this review, we summarize recent advances in data integration-based machine learning for drug response prediction, by categorizing methods as matrix factorization-based, kernel-based and network-based methods. We also present a short description of relevant databases used as a benchmark in drug response prediction analyses, followed by providing a brief discussion of challenges faced in integrating and interpreting data from multiple sources. Finally, we address the advantages of combining multiple heterogeneous data sources on drug sensitivity analysis by showing an experimental comparison. Contact: [email protected].

Recommender systems can support everyday digital tasks by retrieving and recommending useful information contextually. This is becoming increasingly relevant in services and operating systems. Previous research often focuses on specific recommendation tasks with data captured from interactions with an individual application. The quality of recommendations is also often evaluated addressing only computational measures of accuracy, without investigating the usefulness of recommendations in realistic tasks. The aim of this work is to synthesize the research in this area through a novel approach by (1) demonstrating comprehensive digital activity monitoring, (2) introducing entity-based computing and interaction, and (3) investigating the previously overlooked usefulness of entity recommendations and their actual impact on user behavior in real tasks. The methodology exploits context from screen frames recorded every 2 seconds to recommend information entities related to the current task. We embodied this methodology in an interactive system and investigated the relevance and influence of the recommended entities in a study with participants resuming their real-world tasks after a 14-day monitoring phase. Results show that the recommendations allowed participants to find more relevant entities than in a control without the system. In addition, the recommended entities were also used in the actual tasks. In the discussion, we reflect on a research agenda for entity recommendation in context, revisiting comprehensive monitoring to include the physical world, considering entities as actionable recommendations, capturing drifting intent and routines, and considering explainability and transparency of recommendations, ethics, and ownership of data.

In matrix factorization, available graph side-information may not be well suited for the matrix completion problem, having edges that disagree with the latent-feature relations learnt from the incomplete data matrix. We show that removing these contested edges improves prediction accuracy and scalability. We identify the contested edges through a highly-efficient graphical lasso approximation. The identification and removal of contested edges adds no computational complexity to state-of-the-art graph-regularized matrix factorization, remaining linear with respect to the number of non-zeros. Computational load even decreases proportional to the number of edges removed. Formulating a probabilistic generative model and using expectation maximization to extend graph-regularised alternating least squares (GRALS) guarantees convergence. Rich simulated experiments illustrate the desired properties of the resulting algorithm. On real data experiments we demonstrate improved prediction accuracy with fewer graph edges (empirical evidence that graph side-information is often inaccurate). A 300 thousand dimensional graph with three million edges (Yahoo music side-information) can be analyzed in under ten minutes on a standard laptop computer demonstrating the efficiency of our graph update.

We introduce the convolutional spectral kernel (CSK), a novel family of non-stationary, nonparametric covariance kernels for Gaussian process (GP) models, derived from the convolution between two imaginary radial basis functions. We present a principled framework to interpret CSK, as well as other deep probabilistic models, using approximated Fourier transform, yielding a concise representation of input-frequency spectrogram. Observing through the lens of the spectrogram, we provide insight on the interpretability of deep models. We then infer the functional hyperparameters using scalable variational and MCMC methods. On small- and medium-sized spatiotemporal datasets, we demonstrate improved generalization of GP models when equipped with CSK, and their capability to extract non-stationary periodic patterns.

Estimating global pairwise interaction effects, i.e., the difference between the joint effect and the sum of marginal effects of two input features, with uncertainty properly quantified, is centrally important in science applications. We propose a non-parametric probabilistic method for detecting interaction effects of unknown form. First, the relationship between the features and the output is modelled using a Bayesian neural network, capable of representing complex interactions and principled uncertainty. Second, interaction effects and their uncertainty are estimated from the trained model. For the second step, we propose an intuitive global interaction measure: Bayesian Group Expected Hessian (GEH), which aggregates information of local interactions as captured by the Hessian. GEH provides a natural trade-off between type I and type II error and, moreover, comes with theoretical guarantees ensuring that the estimated interaction effects and their uncertainty can be improved by training a more accurate BNN. The method empirically outperforms available non-probabilistic alternatives on simulated and real-world data. Finally, we demonstrate its ability to detect interpretable interactions between higher-level features (at deeper layers of the neural network).

Bayesian optimization is an effective method for finding extrema of a black-box function. We propose a new type of Bayesian optimization for learning user preferences in high-dimensional spaces. The central assumption is that the underlying objective function cannot be evaluated directly, but instead a minimizer along a projection can be queried, which we call a projective preferential query. The form of the query allows for feedback that is natural for a human to give, and which enables interaction. This is demonstrated in a user experiment in which the user feedback comes in the form of optimal position and orientation of a molecule adsorbing to a surface. We demonstrate that our framework is able to find a global minimum of a high-dimensional black-box function, which is an infeasible task for existing preferential Bayesian optimization frameworks that are based on pairwise comparisons.

Graph pooling is a central component of a myriad of graph neural network (GNN) architectures. As an inheritance from traditional CNNs, most approaches formulate graph pooling as a cluster assignment problem, extending the idea of local patches in regular grids to graphs. Despite the wide adherence to this design choice, no work has rigorously evaluated its influence on the success of GNNs. In this paper, we build upon representative GNNs and introduce variants that challenge the need for locality-preserving representations, either using randomization or clustering on the complement graph. Strikingly, our experiments demonstrate that using these variants does not result in any decrease in performance. To understand this phenomenon, we study the interplay between convolutional layers and the subsequent pooling ones. We show that the convolutions play a leading role in the learned representations. In contrast to the common belief, local pooling is not responsible for the success of GNNs on relevant and widely-used benchmarks.

A central concern in an interactive intelligent system is optimization of its actions, to be maximally helpful to its human user. In recommender systems for instance, the action is to choose what to recommend, and the optimization task is to recommend items the user prefers. The optimization is done based on earlier user's feedback (e.g. "likes" and "dislikes"), and the algorithms assume the feedback to be faithful. That is, when the user clicks "like," they actually prefer the item. We argue that this fundamental assumption can be extensively violated by human users, who are not passive feedback sources. Instead, they are in control, actively steering the system towards their goal. To verify this hypothesis, that humans steer and are able to improve performance by steering, we designed a function optimization task where a human and an optimization algorithm collaborate to find the maximum of a 1-dimensional function. At each iteration, the optimization algorithm queries the user for the value of a hidden function f at a point x, and the user, who sees the hidden function, provides an answer about f(x). Our study on 21 participants shows that users who understand how the optimization works, strategically provide biased answers (answers not equal to f(x)), which results in the algorithm finding the optimum significantly faster. Our work highlights that next-generation intelligent systems will need user models capable of helping users who steer systems to pursue their goals.

Abstract The outcome of exploratory information retrieval is not only dependent on the effectiveness of individual responses to a set of queries, but also on relevant information retrieved during the entire exploratory search session. We study the effect of search assistance, operationalized as an interactive faceted query suggestion, for both whole‐session effectiveness and engagement through interactive faceted query suggestion. A user experiment is reported, where users performed exploratory search tasks, comparing interactive faceted query suggestion and a control condition with only conventional typed‐query interaction. Data comprised of interaction and search logs show that the availability of interactive faceted query suggestion substantially improves whole‐session effectiveness by increasing recall without sacrificing precision. The increased engagement with interactive faceted query suggestion is targeted to direct situated navigation around the initial query scope, but is not found to improve individual queries on average. The results imply that research in exploratory search should focus on measuring and designing tools that engage users with directed situated navigation support for improving whole‐session performance.

Abstract A salient approach to interpretable machine learning is to restrict modeling to simple models. In the Bayesian framework, this can be pursued by restricting the model structure and prior to favor interpretable models. Fundamentally, however, interpretability is about users’ preferences, not the data generation mechanism; it is more natural to formulate interpretability as a utility function. In this work, we propose an interpretability utility, which explicates the trade-off between explanation fidelity and interpretability in the Bayesian framework. The method consists of two steps. First, a reference model, possibly a black-box Bayesian predictive model which does not compromise accuracy, is fitted to the training data. Second, a proxy model from an interpretable model family that best mimics the predictive behaviour of the reference model is found by optimizing the interpretability utility function. The approach is model agnostic—neither the interpretable model nor the reference model are restricted to a certain class of models—and the optimization problem can be solved using standard tools. Through experiments on real-word data sets, using decision trees as interpretable models and Bayesian additive regression models as reference models, we show that for the same level of interpretability, our approach generates more accurate models than the alternative of restricting the prior. We also propose a systematic way to measure stability of interpretabile models constructed by different interpretability approaches and show that our proposed approach generates more stable models.

We propose a novel deep learning paradigm of differential flows that learn a stochastic differential equation transformations of inputs prior to a standard classification or regression function. The key property of differential Gaussian processes is the warping of inputs through infinitely deep, but infinitesimal, differential fields, that generalise discrete layers into a dynamical system. We demonstrate excellent results as compared to deep Gaussian processes and Bayesian neural networks.

The expressive power of Gaussian processes depends heavily on the choice of kernel. In this work we propose the novel harmonizable mixture kernel (HMK), a family of expressive, interpretable, non-stationary kernels derived from mixture models on the generalized spectral representation. As a theoretically sound treatment of non-stationary kernels, HMK supports harmonizable covariances, a wide subset of kernels including all stationary and many non-stationary covariances. We also propose variational Fourier features, an inter-domain sparse GP inference framework that offers a representative set of ’inducing frequencies’. We show that harmonizable mixture kernels interpolate between local patterns, and that variational Fourier features offers a robust kernel learning framework for the new kernel family.

Machine learning can help personalized decision support by learning models to predict individual treatment effects (ITE). This work studies the reliability of prediction-based decision-making in a task of deciding which action $a$ to take for a target unit after observing its covariates $\tilde{x}$ and predicted outcomes $\hat{p}(\tilde{y} \mid \tilde{x}, a)$. An example case is personalized medicine and the decision of which treatment to give to a patient. A common problem when learning these models from observational data is imbalance, that is, difference in treated/control covariate distributions, which is known to increase the upper bound of the expected ITE estimation error. We propose to assess the decision-making reliability by estimating the ITE model’s Type S error rate, which is the probability of the model inferring the sign of the treatment effect wrong. Furthermore, we use the estimated reliability as a criterion for active learning, in order to collect new (possibly expensive) observations, instead of making a forced choice based on unreliable predictions. We demonstrate the effectiveness of this decision-making aware active learning in two decision-making tasks: in simulated data with binary outcomes and in a medical dataset with synthetic and continuous treatment outcomes.

Learning predictive models from small high-dimensional data sets is a key problem in high-dimensional statistics. Expert knowledge elicitation can help, and a strong line of work focuses on directly eliciting informative prior distributions for parameters. This either requires considerable statistical expertise or is laborious, as the emphasis has been on accuracy and not on efficiency of the process. Another line of work queries about importance of features one at a time, assuming them to be independent and hence missing covariance information. In contrast, we propose eliciting expert knowledge about pairwise feature similarities, to borrow statistical strength in the predictions, and using sequential decision making techniques to minimize the effort of the expert. Empirical results demonstrate improvement in predictive performance on both simulated and real data, in high-dimensional linear regression tasks, where we learn the covariance structure with a Gaussian process, based on sequential elicitation.

Matrix completion aims to predict missing elements in a partially observed data matrix which in typical applications, such as collaborative filtering, is large and extremely sparsely observed. A standard solution is matrix factorization, which predicts unobserved entries as linear combinations of latent variables. We generalize to non-linear combinations in massive-scale matrices. Bayesian approaches have been proven beneficial in linear matrix completion, but not applied in the more general non-linear case, due to limited scalability. We introduce a Bayesian non-linear matrix completion algorithm, which is based on a recent Bayesian formulation of Gaussian process latent variable models. To solve the challenges regarding scalability and computation, we propose a data-parallel distributed computational approach with a restricted communication scheme. We evaluate our method on challenging out-of-matrix prediction tasks using both simulated and real-world data.

Inferring the relations between two images is an important class of tasks in computer vision. Examples of such tasks include computing optical flow and stereo disparity. We treat the relation inference tasks as a machine learning problem and tackle it with neural networks. A key to the problem is learning a representation of relations. We propose a new neural network module, contrast association unit (CAU), which explicitly models the relations between two sets of input variables. Due to the non-negativity of the weights in CAU, we adopt a multiplicative update algorithm for learning these weights. Experiments show that neural networks with CAUs are more effective in learning five fundamental image transformations than conventional neural networks.

Machine teaching addresses the problem of finding the best training data that can guide a learning algorithm to a target model with minimal effort. In conventional settings, a teacher provides data that are consistent with the true data distribution. However, for sequential learners which actively choose their queries, such as multi-armed bandits and active learners, the teacher can only provide responses to the learner’s queries, not design the full data. In this setting, consistent teachers can be sub-optimal for finite horizons. We formulate this sequential teaching problem, which current techniques in machine teaching do not address, as a Markov decision process, with the dynamics nesting a model of the learner and the actions being the teacher's responses. Furthermore, we address the complementary problem of learning from a teacher that plans: to recognise the teaching intent of the responses, the learner is endowed with a model of the teacher. We test the formulation with multi-armed bandit learners in simulated experiments and a user study. The results show that learning is improved by (i) planning teaching and (ii) the learner having a model of the teacher. The approach gives tools to taking into account strategic (planning) behaviour of users of interactive intelligent systems, such as recommendation engines, by considering them as boundedly optimal teachers.

MOTIVATION: Finding non-linear relationships between biomolecules and a biological outcome is computationally expensive and statistically challenging. Existing methods have important drawbacks, including among others lack of parsimony, non-convexity and computational overhead. Here we propose block HSIC Lasso, a non-linear feature selector that does not present the previous drawbacks. RESULTS: We compare block HSIC Lasso to other state-of-the-art feature selection techniques in both synthetic and real data, including experiments over three common types of genomic data: gene-expression microarrays, single-cell RNA sequencing and genome-wide association studies. In all cases, we observe that features selected by block HSIC Lasso retain more information about the underlying biology than those selected by other techniques. As a proof of concept, we applied block HSIC Lasso to a single-cell RNA sequencing experiment on mouse hippocampus. We discovered that many genes linked in the past to brain development and function are involved in the biological differences between the types of neurons. AVAILABILITY AND IMPLEMENTATION: Block HSIC Lasso is implemented in the Python 2/3 package pyHSICLasso, available on PyPI. Source code is available on GitHub (https://github.com/riken-aip/pyHSICLasso). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Interaction between the genotype and the environment (G×E) has a strong impact on the yield of major crop plants. Although influential, taking G×E explicitly into account in plant breeding has remained difficult. Recently G×E has been predicted from environmental and genomic covariates, but existing works have not shown that generalization to new environments and years without access to in-season data is possible and practical applicability remains unclear. Using data from a Barley breeding programme in Finland, we construct an in silico experiment to study the viability of G×E prediction under practical constraints. RESULTS: We show that the response to the environment of a new generation of untested Barley cultivars can be predicted in new locations and years using genomic data, machine learning and historical weather observations for the new locations. Our results highlight the need for models of G×E: non-linear effects clearly dominate linear ones, and the interaction between the soil type and daily rain is identified as the main driver for G×E for Barley in Finland. Our study implies that genomic selection can be used to capture the yield potential in G×E effects for future growth seasons, providing a possible means to achieve yield improvements, needed for feeding the growing population. AVAILABILITY AND IMPLEMENTATION: The data accompanied by the method code (http://research.cs.aalto.fi/pml/software/gxe/bioinformatics_codes.zip) is available in the form of kernels to allow reproducing the results. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Metabolic flux balance analysis (FBA) is a standard tool in analyzing metabolic reaction rates compatible with measurements, steady-state and the metabolic reaction network stoichiometry. Flux analysis methods commonly place model assumptions on fluxes due to the convenience of formulating the problem as a linear programing model, while many methods do not consider the inherent uncertainty in flux estimates. RESULTS: We introduce a novel paradigm of Bayesian metabolic flux analysis that models the reactions of the whole genome-scale cellular system in probabilistic terms, and can infer the full flux vector distribution of genome-scale metabolic systems based on exchange and intracellular (e.g. 13C) flux measurements, steady-state assumptions, and objective function assumptions. The Bayesian model couples all fluxes jointly together in a simple truncated multivariate posterior distribution, which reveals informative flux couplings. Our model is a plug-in replacement to conventional metabolic balance methods, such as FBA. Our experiments indicate that we can characterize the genome-scale flux covariances, reveal flux couplings, and determine more intracellular unobserved fluxes in Clostridium acetobutylicum from 13C data than flux variability analysis. AVAILABILITY AND IMPLEMENTATION: The COBRA compatible software is available at github.com/markusheinonen/bamfa. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Human genomic datasets often contain sensitive information that limits use and sharing of the data. In particular, simple anonymization strategies fail to provide sufficient level of protection for genomic data, because the data are inherently identifiable. Differentially private machine learning can help by guaranteeing that the published results do not leak too much information about any individual data point. Recent research has reached promising results on differentially private drug sensitivity prediction using gene expression data. Differentially private learning with genomic data is challenging because it is more difficult to guarantee privacy in high dimensions. Dimensionality reduction can help, but if the dimension reduction mapping is learned from the data, then it needs to be differentially private too, which can carry a significant privacy cost. Furthermore, the selection of any hyperparameters (such as the target dimensionality) needs to also avoid leaking private information. RESULTS: We study an approach that uses a large public dataset of similar type to learn a compact representation for differentially private learning. We compare three representation learning methods: variational autoencoders, principal component analysis and random projection. We solve two machine learning tasks on gene expression of cancer cell lines: cancer type classification, and drug sensitivity prediction. The experiments demonstrate significant benefit from all representation learning methods with variational autoencoders providing the most accurate predictions most often. Our results significantly improve over previous state-of-the-art in accuracy of differentially private drug sensitivity prediction. AVAILABILITY AND IMPLEMENTATION: Code used in the experiments is available at https://github.com/DPBayes/dp-representation-transfer.

The use of implicit relevance feedback from neurophysiology could deliver effortless information retrieval. However, both computing neurophysiologic responses and retrieving documents are characterized by uncertainty because of noisy signals and incomplete or inconsistent representations of the data. We present the first-of-its-kind, fully integrated information retrieval system that makes use of online implicit relevance feedback generated from brain activity as measured through electroencephalography (EEG), and eye movements. The findings of the evaluation experiment (N = 16) show that we are able to compute online neurophysiology-based relevance feedback with performance significantly better than chance in complex data domains and realistic search tasks. We contribute by demonstrating how to integrate in interactive intent modeling this inherently noisy implicit relevance feedback combined with scarce explicit feedback. Although experimental measures of task performance did not allow us to demonstrate how the classification outcomes translated into search task performance, the experiment proved that our approach is able to generate relevance feedback from brain signals and eye movements in a realistic scenario, thus providing promising implications for future work in neuroadaptive information retrieval (IR).

Bayesian matrix factorization (BMF) is a powerful tool for producing low-rank representations of matrices and for predicting missing values and providing confidence intervals. Scaling up the posterior inference for massive-scale matrices is challenging and requires distributing both data and computation over many workers, making communication the main computational bottleneck. Embarrassingly parallel inference would remove the communication needed, by using completely independent computations on different data subsets, but it suffers from the inherent unidentifiability of BMF solutions. We introduce a hierarchical decomposition of the joint posterior distribution, which couples the subset inferences, allowing for embarrassingly parallel computations in a sequence of at most three stages. Using an efficient approximate implementation, we show improvements empirically on both real and simulated data. Our distributed approach is able to achieve a speed-up of almost an order of magnitude over the full posterior, with a negligible effect on predictive accuracy. Our method outperforms state-of-the-art embarrassingly parallel MCMC methods in accuracy, and achieves results competitive to other available distributed and parallel implementations of BMF.

In human-in-the-loop machine learning, the user provides information beyond that in the training data. Many algorithms and user interfaces have been designed to optimize and facilitate this human--machine interaction; however, fewer studies have addressed the potential defects the designs can cause. Effective interaction often requires exposing the user to the training data or its statistics. The design of the system is then critical, as this can lead to double use of data and overfitting, if the user reinforces noisy patterns in the data. We propose a user modelling methodology, by assuming simple rational behaviour, to correct the problem. We show, in a user study with 48 participants, that the method improves predictive performance in a sparse linear regression sentiment analysis task, where graded user knowledge on feature relevance is elicited. We believe that the key idea of inferring user knowledge with probabilistic user models has general applicability in guarding against overfitting and improving interactive machine learning.

Motivation: Precision medicine requires the ability to predict the efficacies of different treatments for a given individual using high-dimensional genomic measurements. However, identifying predictive features remains a challenge when the sample size is small. Incorporating expert knowledge offers a promising approach to improve predictions, but collecting such knowledge is laborious if the number of candidate features is very large. Results: We introduce a probabilistic framework to incorporate expert feedback about the impact of genomic measurements on the outcome of interest and present a novel approach to collect the feedback efficiently, based on Bayesian experimental design. The new approach outperformed other recent alternatives in two medical applications: prediction of metabolic traits and prediction of sensitivity of cancer cells to different drugs, both using genomic features as predictors. Furthermore, the intelligent approach to collect feedback reduced the workload of the expert to approximately 11%, compared to a baseline approach. Availability and implementation: Source code implementing the introduced computational methods is freely available at https://github.com/AaltoPML/knowledge-elicitation-for-precision-medicine. Supplementary information: Supplementary data are available at Bioinformatics online.

Engine for Likelihood-Free Inference (ELFI) is a Python software library for performing likelihood-free inference (LFI). ELFI provides a convenient syntax for arranging components in LFI, such as priors, simulators, summaries or distances, to a network called ELFI graph. The components can be implemented in a wide variety of languages. The stand-alone ELFI graph can be used with any of the available inference methods without modifications. A central method implemented in ELFI is Bayesian Optimization for Likelihood-Free Inference (BOLFI), which has recently been shown to accelerate likelihood-free inference up to several orders of magnitude by surrogate-modelling the distance. ELFI also has an inbuilt support for output data storing for reuse and analysis, and supports parallelization of computation from multiple cores up to a cluster environment. ELFI is designed to be extensible and provides interfaces for widening its functionality. This makes the adding of new inference methods to ELFI straightforward and automatically compatible with the inbuilt features.

Exploratory search requires the system to assist the user in comprehending the information space and expressing evolving search intents for iterative exploration and retrieval of information. We introduce interactive intent modeling, a technique that models a user’s evolving search intents and visualizes them as keywords for interaction. The user can provide feedback on the keywords, from which the system learns and visualizes an improved intent estimate and retrieves information. We report experiments comparing variants of a system implementing interactive intent modeling to a control system. Data comprising search logs, interaction logs, essay answers, and questionnaires indicate significant improvements in task performance, information retrieval performance over the session, information comprehension performance, and user experience. The improvements in retrieval effectiveness can be attributed to the intent modeling and the effect on users’ task performance, breadth of information comprehension, and user experience are shown to be dependent on a richer visualization. Our results demonstrate the utility of combining interactive modeling of search intentions with interactive visualization of the models that can benefit both directing the exploratory search process and making sense of the information space. Our findings can help design personalized systems that support exploratory information seeking and discovery of novel information.

We introduce a novel kernel that models input-dependent couplings across multiple latent processes. The pairwise joint kernel measures covariance along inputs and across different latent signals in a mutually-dependent fashion. A latent correlation Gaussian process (LCGP) model combines these non-stationary latent components into multiple outputs by an input-dependent mixing matrix. Probit classification and support for multiple observation sets are derived by Variational Bayesian inference. Results on several datasets indicate that the LCGP model can recover the correlations between latent signals while simultaneously achieving state-of-the-art performance. We highlight the latent covariances with an EEG classification dataset where latent brain processes and their couplings simultaneously emerge from the model.

We introduce the localized Lasso, which learns models that both are interpretable and have a high predictive power in problems with high dimensionality d and small sample size n. More specifically, we consider a function defined by local sparse models, one at each data point. We introduce sample-wise network regularization to borrow strength across the models, and sample-wise exclusive group sparsity (a.k.a., l12 norm) to introduce diversity into the choice of feature sets in the local models. The local models are interpretable in terms of similarity of their sparsity patterns. The cost function is convex, and thus has a globally optimal solution. Moreover, we propose a simple yet efficient iterative least-squares based optimization procedure for the localized Lasso, which does not need a tuning parameter, and is guaranteed to converge to a globally optimal solution. The solution is empirically shown to outperform alternatives for both simulated and genomic personalized/precision medicine data.

An important problem for HCI researchers is to estimate the parameter values of a cognitive model from behavioral data. This is a difficult problem, because of the substantial complexity and variety in human behavioral strategies. We report an investigation into a new approach using approximate Bayesian computation (ABC) to condition model parameters to data and prior knowledge. As the case study we examine menu interaction, where we have click time data only to infer a cognitive model that implements a search behaviour with parameters such as fixation duration and recall probability. Our results demonstrate that ABC (i) improves estimates of model parameter values, (ii) enables meaningful comparisons between model variants, and (iii) supports fitting models to individual users. ABC provides ample opportunities for theoretical HCI research by allowing principled inference of model parameter values and their uncertainty.

Providing accurate predictions is challenging for machine learning algorithms when the number of features is larger than the number of samples in the data. Prior knowledge can improve machine learning models by indicating relevant variables and parameter values. Yet, this prior knowledge is often tacit and only available from domain experts. We present a novel approach that uses interactive visualization to elicit the tacit prior knowledge and uses it to improve the accuracy of prediction models. The main component of our approach is a user model that models the domain expert's knowledge of the relevance of different features for a prediction task. In particular, based on the expert's earlier input, the user model guides the selection of the features on which to elicit user's knowledge next. The results of a controlled user study show that the user model significantly improves prior knowledge elicitation and prediction accuracy, when predicting the relative citation counts of scientific documents in a specific domain.

We introduce the convex factorization machine (CFM), which is a convex variant of the widely used Factorization Machines (FMs). Specifically, we employ a linear+quadratic model and regularize the linear term with the ℓ2-regularizer and the quadratic term with the trace norm regularizer. Then, we formulate the CFM optimization as a semidefinite programming problem and propose an efficient optimization procedure with Hazan's algorithm. A key advantage of CFM over existing FMs is that it can find a globally optimal solution, while FMs may get a poor locally optimal solution since the objective function of FMs is non-convex. In addition, the proposed algorithm is simple yet effective and can be implemented easily. Finally, CFM is a general factorization method and can also be used for other factorization problems, including multi-view matrix factorization and tensor completion problems, in various domains including toxicogenomics and bioinformatics. Through synthetic and traditionally used movielens datasets, we first show that the proposed CFM achieves results competitive to FMs. We then show in a toxicogenomics prediction task that CFM predicts the toxic outcomes of a collection of drugs better than a state-of-the-art tensor factorization method.

Many applications of machine learning, for example in health care, would benefit from methods that can guarantee privacy of data subjects. Differential privacy (DP) has become established as a standard for protecting learning results. The standard DP algorithms require a single trusted party to have access to the entire data, which is a clear weakness, or add prohibitive amounts of noise. We consider DP Bayesian learning in a distributed setting, where each party only holds a single sample or a few samples of the data. We propose a learning strategy based on a secure multi-party sum function for aggregating summaries from data holders and the Gaussian mechanism for DP. Our method builds on an asymptotically optimal and practically efficient DP Bayesian inference with rapidly diminishing extra cost.

We propose non-stationary spectral kernels for Gaussian process regression by modelling the spectral density of a non-stationary kernel function as a mixture of input-dependent Gaussian process frequency density surfaces. We solve the generalised Fourier transform with such a model, and present a family of non-stationary and non-monotonic kernels that can learn input-dependent and potentially long-range, non-monotonic covariances between inputs. We derive efficient inference using model whitening and marginalized posterior, and show with case studies that these kernels are necessary when modelling even rather simple time series, image or geospatial data with non-stationary characteristics.

Regression under the "small n$, large p" condition, of small sample size n and large number of features p in the learning data set, is a recurring setting in which learning from data is difficult. With prior knowledge about relationships of the features, p can effectively be reduced, but explicating such prior knowledge is difficult for experts. In this paper we introduce a new method for eliciting expert prior knowledge about the similarity of the roles of features in the prediction task. The key idea is to use an interactive multidimensional-scaling (MDS) type scatterplot display of the features to elicit the similarity relationships, and then use the elicited relationships in the prior distribution of prediction parameters. Specifically, for learning to predict a target variable with Bayesian linear regression, the feature relationships are used to construct a Gaussian prior with a full covariance matrix for the regression coefficients. Evaluation of our method in experiments with simulated and real users on text data confirm that prior elicitation of feature similarities improves prediction accuracy. Furthermore, elicitation with an interactive scatterplot display outperforms straightforward elicitation where the users choose feature pairs from a feature list.

BACKGROUND: Dispersed biomedical databases limit user exploration to generate structured knowledge. Linked Data unifies data structures and makes the dispersed data easy to search across resources, but it lacks supporting human cognition to achieve insights. In addition, potential errors in the data are difficult to detect in their free formats. Devising a visualization that synthesizes multiple sources in such a way that links between data sources are transparent, and uncertainties, such as data conflicts, are salient is challenging. RESULTS: To investigate the requirements and challenges of uncertainty-aware visualizations of linked data, we developed MediSyn, a system that synthesizes medical datasets to support drug treatment selection. It uses a matrix-based layout to visually link drugs, targets (e.g., mutations), and tumor types. Data uncertainties are salient in MediSyn; for example, (i) missing data are exposed in the matrix view of drug-target relations; (ii) inconsistencies between datasets are shown via overlaid layers; and (iii) data credibility is conveyed through links to data provenance. CONCLUSIONS: Through the synthesis of two manually curated datasets, cancer treatment biomarkers and drug-target bioactivities, a use case shows how MediSyn effectively supports the discovery of drug-repurposing opportunities. A study with six domain experts indicated that MediSyn benefited the drug selection and data inconsistency discovery. Though linked publication sources supported user exploration for further information, the causes of inconsistencies were not easy to find. Additionally, MediSyn could embrace more patient data to increase its informativeness. We derive design implications from the findings.

The R package GFA provides a full pipeline for factor analysis of multiple data sources that are represented as matrices with co-occurring samples. It allows learning dependencies between subsets of the data sources, decomposed into latent factors. The package also implements sparse priors for the factorization, providing interpretable biclusters of the multi-source data.

Prediction in a small-sized sample with a large number of covariates, the “small n , large p ” problem, is challenging. This setting is encountered in multiple applications, such as in precision medicine, where obtaining additional data can be extremely costly or even impossible, and extensive research effort has recently been dedicated to finding principled solutions for accurate prediction. However, a valuable source of additional information, domain experts, has not yet been efficiently exploited. We formulate knowledge elicitation generally as a probabilistic inference process, where expert knowledge is sequentially queried to improve predictions. In the specific case of sparse linear regression, where we assume the expert has knowledge about the relevance of the covariates, or of values of the regression coefficients, we propose an algorithm and computational approximation for fast and efficient interaction, which sequentially identifies the most informative features on which to query expert knowledge. Evaluations of the proposed method in experiments with simulated and real users show improved prediction accuracy already with a small effort from the expert.

We present a novel approach for non-stationary Gaussian process regression (GPR), where the three key parameters – noise variance, signal variance and lengthscale – can be simultaneously input-dependent. We develop gradient-based inference methods to learn the unknown function and the non-stationary model parameters, without requiring any model approximations. For inferring the full posterior distribution we use Hamiltonian Monte Carlo (HMC), which conveniently extends the analytical gradient-based GPR learning by guiding the sampling with the gradients. The MAP solution can also be learned with gradient ascent. In experiments on several synthetic datasets and in modelling of temporal gene expression, the non-stationary GPR is shown to give major improvement when modeling realistic input-dependent dynamics.

A main goal of data visualization is to find, from among all the available alternatives, mappings to the 2D/3D display which are relevant to the user. Assuming user interaction data, or other auxiliary data about the items or their relationships, the goal is to identify which aspects in the primary data support the user's input and, equally importantly, which aspects of the user's potentially noisy input have support in the primary data. For solving the problem, we introduce a multi-view embedding in which a latent factorization identifies which aspects in the two data views (primary data and user data) are related and which are specific to only one of them. The factorization is a generative model in which the display is parameterized as a part of the factorization and the other factors explain away the aspects not expressible in a two-dimensional display. Functioning of the model is demonstrated on several data sets.

We consider regression under the "extremely small n large p" condition, where the number of samples n is so small compared to the dimensionality p that predictors cannot be estimated without prior knowledge. This setup occurs in personalized medicine, for instance, when predicting treatment outcomes for an individual patient based on noisy high-dimensional genomics data. A remaining source of information is expert knowledge, which has received relatively little attention in recent years. We formulate the inference problem of asking expert feedback on features on a budget, propose an elicitation strategy for a simple "small n" setting, and derive conditions under which the elicitation strategy is optimal. Experiments on simulated experts, both on synthetic and genomics data, demonstrate that the proposed strategy can drastically improve prediction accuracy.

In exploratory search, the user starts with an uncertain information need and provides relevance feedback to the system's suggestions to direct the search. The search system learns the user intent based on this feedback and employs it to recommend novel results. However, the amount of user feedback is very limited compared to the size of the information space to be explored. To tackle this problem, we take into account user feedback on both the retrieved items (documents) and their features (keywords). In order to combine feedback from multiple domains, we introduce a coupled multi-armed bandits algorithm, which employs a probabilistic model of the relationship between the domains. Simulation results show that with multi-domain feedback, the search system can find the relevant items in fewer iterations than with only one domain. A preliminary user study indicates improvement in user satisfaction and quality of retrieved information.

In exploratory search tasks, users usually start with considerable uncertainty about their search goals, and so the search intent of the user may be volatile as the user is constantly learning and reformulating her search hypothesis during the search. This may lead to a noticeable concept drift in the relevance feedback given by the user. We formulate a Bayesian regression model for predicting the accuracy of each individual user feedback and thus find outliers in the feedback data set. To accompany this model, we introduce a timeline interface that visualizes the feedback history to the user and gives her suggestions on which past feedback is likely in need of adjustment. This interface also allows the user to adjust the feedback accuracy inferences made by the model. Simulation experiments demonstrate that the performance of the new user model outperforms a simpler baseline and that the performance approaches that of an oracle, given a small amount of additional user interaction. A user study shows that the proposed modeling technique, combined with the timeline interface, made it easier for the users to notice and correct mistakes in their feedback, resulted in better and more diverse recommendations, allowed users to easier find items they liked, and was more understandable.

MOTIVATION: A key goal of computational personalized medicine is to systematically utilize genomic and other molecular features of samples to predict drug responses for a previously unseen sample. Such predictions are valuable for developing hypotheses for selecting therapies tailored for individual patients. This is especially valuable in oncology, where molecular and genetic heterogeneity of the cells has a major impact on the response. However, the prediction task is extremely challenging, raising the need for methods that can effectively model and predict drug responses. RESULTS: In this study, we propose a novel formulation of multi-task matrix factorization that allows selective data integration for predicting drug responses. To solve the modeling task, we extend the state-of-the-art kernelized Bayesian matrix factorization (KBMF) method with component-wise multiple kernel learning. In addition, our approach exploits the known pathway information in a novel and biologically meaningful fashion to learn the drug response associations. Our method quantitatively outperforms the state of the art on predicting drug responses in two publicly available cancer datasets as well as on a synthetic dataset. In addition, we validated our model predictions with lab experiments using an in-house cancer cell line panel. We finally show the practical applicability of the proposed method by utilizing prior knowledge to infer pathway-drug response associations, opening up the opportunity for elucidating drug action mechanisms. We demonstrate that pathway-response associations can be learned by the proposed model for the well-known EGFR and MEK inhibitors. AVAILABILITY AND IMPLEMENTATION: The source code implementing the method is available at http://research.cs.aalto.fi/pml/software/cwkbmf/ CONTACTS: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Public and private repositories of experimental data are growing to sizes that require dedicated methods for finding relevant data. To improve on the state of the art of keyword searches from annotations, methods for content-based retrieval have been proposed. In the context of gene expression experiments, most methods retrieve gene expression profiles, requiring each experiment to be expressed as a single profile, typically of case versus control. A more general, recently suggested alternative is to retrieve experiments whose models are good for modelling the query dataset. However, for very noisy and high-dimensional query data, this retrieval criterion turns out to be very noisy as well. RESULTS: We propose doing retrieval using a denoised model of the query dataset, instead of the original noisy dataset itself. To this end, we introduce a general probabilistic framework, where each experiment is modelled separately and the retrieval is done by finding related models. For retrieval of gene expression experiments, we use a probabilistic model called product partition model, which induces a clustering of genes that show similar expression patterns across a number of samples. The suggested metric for retrieval using clusterings is the normalized information distance. Empirical results finally suggest that inference for the full probabilistic model can be approximated with good performance using computationally faster heuristic clustering approaches (e.g. k-means). The method is highly scalable and straightforward to apply to construct a general-purpose gene expression experiment retrieval method. AVAILABILITY AND IMPLEMENTATION: The method can be implemented using standard clustering algorithms and normalized information distance, available in many statistical software packages. CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Modelling methods that find structure in data are necessary with the current large volumes of genomic data, and there have been various efforts to find subsets of genes exhibiting consistent patterns over subsets of treatments. These biclustering techniques have focused on one data source, often gene expression data. We present a Bayesian approach for joint biclustering of multiple data sources, extending a recent method Group Factor Analysis to have a biclustering interpretation with additional sparsity assumptions. The resulting method enables data-driven detection of linear structure present in parts of the data sources. RESULTS: Our simulation studies show that the proposed method reliably infers biclusters from heterogeneous data sources. We tested the method on data from the NCI-DREAM drug sensitivity prediction challenge, resulting in an excellent prediction accuracy. Moreover, the predictions are based on several biclusters which provide insight into the data sources, in this case on gene expression, DNA methylation, protein abundance, exome sequence, functional connectivity fingerprints and drug sensitivity. AVAILABILITY AND IMPLEMENTATION: http://research.cs.aalto.fi/pml/software/GFAsparse/ CONTACTS: : [email protected] or [email protected].

In high-dimensional data, structured noise caused by observed and unobserved factors affecting multiple target variables simultaneously, imposes a serious challenge for modeling, by masking the often weak signal. Therefore, (1) explaining away the structured noise in multiple-output regression is of paramount importance. Additionally, (2) assumptions about the correlation structure of the regression weights are needed. We note that both can be formulated in a natural way in a latent variable model, in which both the interesting signal and the noise are mediated through the same latent factors. Under this assumption, the signal model then borrows strength from the noise model by encouraging similar effects on correlated targets. We introduce a hyperparameter for the latent signal-to-noise ratio which turns out to be important for modelling weak signals, and an ordered infinite-dimensional shrinkage prior that resolves the rotational unidentifiability in reduced-rank regression models. Simulations and prediction experiments with metabolite, gene expression, FMRI measurement, and macroeconomic time series data show that our model equals or exceeds the state-of-the-art performance and, in particular, outperforms the standard approach of assuming independent noise and signal models.

Nonlinear dimensionality reduction by manifold embedding has become a popular and powerful approach both for visualization and as preprocessing for predictive tasks, but more efficient optimization algorithms are still crucially needed. Majorization-Minimization (MM) is a promising approach that monotonically decreases the cost function, but it remains unknown how to tightly majorize the manifold embedding objective functions such that the resulting MM algorithms are efficient and robust. We propose a new MM procedure that yields fast MM algorithms for a wide variety of manifold embedding problems. In our majorization step, two parts of the cost function are respectively upper bounded by quadratic and Lipschitz surrogates, and the resulting upper bound can be minimized in closed form. For cost functions amenable to such QL-majorization, the MM yields monotonic improvement and is efficient: in experiments the newly developed MM algorithms outperform five state-of-the-art optimization approaches in manifold embedding tasks.

Peripheral physiological signals, as obtained using electrodermal activity and facial electromyography over the corrugator supercilii muscle, are explored as indicators of perceived relevance in information retrieval tasks. An experiment with 40 participants is reported, in which these physiological signals are recorded while participants perform information retrieval tasks. Appropriate feature engineering is defined, and the feature space is explored. The results indicate that features in the window of 4 to 6 seconds after the relevance judgment for electrodermal activity, and from 1 second before to 2 seconds after the relevance judgment for corrugator supercilii activity, are associated with the users' perceived relevance of information items. A classifier verified the predictive power of the features and showed up to 14% improvement predicting relevance. Our research can help the design of intelligent user interfaces for information retrieval that can detect the user's perceived relevance from physiological signals and complement or replace conventional relevance feedback.

In exploratory search, when a user directs a search engine using uncertain relevance feedback, usability problems regarding controllability and predictability may arise. One problem is that the user is often modelled as a passive source of relevance information, instead of an active entity trying to steer the system based on evolving information needs. This may cause the user to feel that the response of the system is inconsistent with her steering. Another problem arises due to the sheer size and complexity of the information space, and hence of the system, as it may be difficult for the user to anticipate the consequences of her actions in this complex environment. These problems can be mitigated by interpreting the user's actions as setting a goal for an optimization problem regarding the system state, instead of passive relevance feedback, and by allowing the user to see the predicted effects of an action before committing to it. In this paper, we present an implementation of these improvements in a visual user-controllable search interface. A user study involving exploratory search for scientific literature gives some indication on improvements in task performance, usability, perceived usefulness and user acceptance.

Current search engines offer limited assistance for exploration and information discovery in complex search tasks. Instead, users are distracted by the need to focus their cognitive efforts on finding navigation cues, rather than selecting relevant information. Interactive intent modeling enhances the human information exploration capacity through computational modeling, visualized for interaction. Interactive intent modeling has been shown to increase task-level information seeking performance by up to 100%. In this demonstration, we showcase SciNet, a system implementing interactive intent modeling on top of a scientific article database of over 60 million documents.

Factor analysis (FA) provides linear factors that describe the relationships between individual variables of a data set. We extend this classical formulation into linear factors that describe the relationships between groups of variables, where each group represents either a set of related variables or a data set. The model also naturally extends canonical correlation analysis to more than two sets, in a way that is more flexible than previous extensions. Our solution is formulated as a variational inference of a latent variable model with structural sparsity, and it consists of two hierarchical levels: 1) the higher level models the relationships between the groups and 2) the lower models the observed variables given the higher level. We show that the resulting solution solves the group factor analysis (GFA) problem accurately, outperforming alternative FA-based solutions as well as more straightforward implementations of GFA. The method is demonstrated on two life science data sets, one on brain activation and the other on systems biology, illustrating its applicability to the analysis of different types of high-dimensional data sources.

We present a novel approach to low-dimensional neighbor embedding for visualization, based on formulating an information retrieval based neighborhood preservation cost function as Maximum satisfiability on a discretized output display. The method has a rigorous interpretation as optimal visualization based on the cost function. Unlike previous low-dimensional neighbor embedding methods, our formulation is guaranteed to yield globally optimal visualizations, and does so reasonably fast. Unlike previous manifold learning methods yielding global optima of their cost functions, our cost function and method are designed for low-dimensional visualization where evaluation and minimization of visualization errors are crucial. Our method performs well in experiments, yielding clean embeddings of datasets where a state-of-the-art comparison method yields poor arrangements. In a real-world case study for semi-supervised WLAN signal mapping in buildings we outperform state-of-the-art methods.

Visualization methods that arrange data objects in 2D or 3D layouts have followed two main schools, methods oriented for graph layout and methods oriented for vectorial embedding. We show the two previously separate approaches are tied by an optimization equivalence, making it possible to relate methods from the two approaches and to build new methods that take the best of both worlds. In detail, we prove a theorem of optimization equivalences between beta- and gamma-, as well as alpha- and Renyi-divergences through a connection scalar. Through the equivalences we represent several nonlinear dimensionality reduction and graph drawing methods in a generalized stochastic neighbor embedding setting, where information divergences are minimized between similarities in input and output spaces, and the optimal connection scalar provides a natural choice for the tradeoff between attractive and repulsive forces. We give two examples of developing new visualization methods through the equivalences: 1) We develop weighted symmetric stochastic neighbor embedding (ws-SNE) from Elastic Embedding and analyze its benefits, good performance for both vectorial and network data; in experiments ws-SNE has good performance across data sets of different types, whereas comparison methods fail for some of the data sets; 2) we develop a gamma-divergence version of a PolyLog layout method; the new method is scale invariant in the output space and makes it possible to efficiently use large-scale smoothed neighborhoods.

Term-Relevance Prediction from Brain Signals (TRPB) is proposed to automatically detect relevance of text information directly from brain signals. An experiment with forty participants was conducted to record neural activity of participants while providing relevance judgments to text stimuli for a given topic. High-precision scientific equipment was used to quantify neural activity across 32 electroencephalography (EEG) channels. A classifier based on a multi-view EEG feature representation showed improvement up to 17% in relevance prediction based on brain signals alone. Relevance was also associated with brain activity with significant changes in certain brain areas. Consequently, TRPB is based on changes identified in specific brain areas and does not require user-specific training or calibration. Hence, relevance predictions can be conducted for unseen content and unseen participants. As an application of TRPB we demonstrate a high-precision variant of the classifier that constructs sets of relevant terms for a given unknown topic of interest. Our research shows that detecting relevance from brain signals is possible and allows the acquisition of relevance judgments without a need to observe any other user interaction. This suggests that TRPB could be used in combination or as an alternative for conventional implicit feedback signals, such as dwell time or click-through activity.

MOTIVATION: Analysis of relationships of drug structure to biological response is key to understanding off-target and unexpected drug effects, and for developing hypotheses on how to tailor drug therapies. New methods are required for integrated analyses of a large number of chemical features of drugs against the corresponding genome-wide responses of multiple cell models. RESULTS: In this article, we present the first comprehensive multi-set analysis on how the chemical structure of drugs impacts on genome-wide gene expression across several cancer cell lines [Connectivity Map (CMap) database]. The task is formulated as searching for drug response components across multiple cancers to reveal shared effects of drugs and the chemical features that may be responsible. The components can be computed with an extension of a recent approach called Group Factor Analysis. We identify 11 components that link the structural descriptors of drugs with specific gene expression responses observed in the three cell lines and identify structural groups that may be responsible for the responses. Our method quantitatively outperforms the limited earlier methods on CMap and identifies both the previously reported associations and several interesting novel findings, by taking into account multiple cell lines and advanced 3D structural descriptors. The novel observations include: previously unknown similarities in the effects induced by 15-delta prostaglandin J2 and HSP90 inhibitors, which are linked to the 3D descriptors of the drugs; and the induction by simvastatin of leukemia-specific response, resembling the effects of corticosteroids. AVAILABILITY AND IMPLEMENTATION: Source Code implementing the method is available at: http://research.ics.aalto.fi/mi/software/GFAsparse. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: A typical genome-wide association study searches for associations between single nucleotide polymorphisms (SNPs) and a univariate phenotype. However, there is a growing interest to investigate associations between genomics data and multivariate phenotypes, for example, in gene expression or metabolomics studies. A common approach is to perform a univariate test between each genotype-phenotype pair, and then to apply a stringent significance cutoff to account for the large number of tests performed. However, this approach has limited ability to uncover dependencies involving multiple variables. Another trend in the current genetics is the investigation of the impact of rare variants on the phenotype, where the standard methods often fail owing to lack of power when the minor allele is present in only a limited number of individuals. RESULTS: We propose a new statistical approach based on Bayesian reduced rank regression to assess the impact of multiple SNPs on a high-dimensional phenotype. Because of the method's ability to combine information over multiple SNPs and phenotypes, it is particularly suitable for detecting associations involving rare variants. We demonstrate the potential of our method and compare it with alternatives using the Northern Finland Birth Cohort with 4702 individuals, for whom genome-wide SNP data along with lipoprotein profiles comprising 74 traits are available. We discovered two genes (XRCC4 and MTHFD2L) without previously reported associations, which replicated in a combined analysis of two additional cohorts: 2390 individuals from the Cardiovascular Risk in Young Finns study and 3659 individuals from the FINRISK study. AVAILABILITY AND IMPLEMENTATION: R-code freely available for download at http://users.ics.aalto.fi/pemartti/gene_metabolome/.

MOTIVATION: Over the recent years, the field of whole-metagenome shotgun sequencing has witnessed significant growth owing to the high-throughput sequencing technologies that allow sequencing genomic samples cheaper, faster and with better coverage than before. This technical advancement has initiated the trend of sequencing multiple samples in different conditions or environments to explore the similarities and dissimilarities of the microbial communities. Examples include the human microbiome project and various studies of the human intestinal tract. With the availability of ever larger databases of such measurements, finding samples similar to a given query sample is becoming a central operation. RESULTS: In this article, we develop a content-based exploration and retrieval method for whole-metagenome sequencing samples. We apply a distributed string mining framework to efficiently extract all informative sequence k-mers from a pool of metagenomic samples and use them to measure the dissimilarity between two samples. We evaluate the performance of the proposed approach on two human gut metagenome datasets as well as human microbiome project metagenomic samples. We observe significant enrichment for diseased gut samples in results of queries with another diseased sample and high accuracy in discriminating between different body sites even though the method is unsupervised. AVAILABILITY AND IMPLEMENTATION: A software implementation of the DSM framework is available at https://github.com/HIITMetagenomics/dsm-framework. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Data analysis for metabolomics suffers from uncertainty because of the noisy measurement technology and the small sample size of experiments. Noise and the small sample size lead to a high probability of false findings. Further, individual compounds have natural variation between samples, which in many cases renders them unreliable as biomarkers. However, the levels of similar compounds are typically highly correlated, which is a phenomenon that we model in this work. RESULTS: We propose a hierarchical Bayesian model for inferring differences between groups of samples more accurately in metabolomic studies, where the observed compounds are collinear. We discover that the method decreases the error of weak and non-existent covariate effects, and thereby reduces false-positive findings. To achieve this, the method makes use of the mass spectral peak data by clustering similar peaks into latent compounds, and by further clustering latent compounds into groups that respond in a coherent way to the experimental covariates. We demonstrate the method with three simulated studies and validate it with a metabolomic benchmark dataset. AVAILABILITY AND IMPLEMENTATION: An implementation in R is available at http://research.ics.aalto.fi/mi/software/peakANOVA/.

BACKGROUND: The aim of connectivity mapping is to match drugs using drug-treatment gene expression profiles from multiple cell lines. This can be viewed as an information retrieval task, with the goal of finding the most relevant profiles for a given query drug. We infer the relevance for retrieval by data-driven probabilistic modeling of the drug responses, resulting in probabilistic connectivity mapping, and further consider the available cell lines as different data sources. We use a special type of probabilistic model to separate what is shared and specific between the sources, in contrast to earlier connectivity mapping methods that have intentionally aggregated all available data, neglecting information about the differences between the cell lines. RESULTS: We show that the probabilistic multi-source connectivity mapping method is superior to alternatives in finding functionally and chemically similar drugs from the Connectivity Map data set. We also demonstrate that an extension of the method is capable of retrieving combinations of drugs that match different relevant parts of the query drug response profile. CONCLUSIONS: The probabilistic modeling-based connectivity mapping method provides a promising alternative to earlier methods. Principled integration of data from different cell lines helps to identify relevant responses for specific drug repositioning applications.

BACKGROUND: Mass spectrometry-based metabolomic analysis depends upon the identification of spectral peaks by their mass and retention time. Statistical analysis that follows the identification currently relies on one main peak of each compound. However, a compound present in the sample typically produces several spectral peaks due to its isotopic properties and the ionization process of the mass spectrometer device. In this work, we investigate the extent to which these additional peaks can be used to increase the statistical strength of differential analysis. RESULTS: We present a Bayesian approach for integrating data of multiple detected peaks that come from one compound. We demonstrate the approach through a simulated experiment and validate it on ultra performance liquid chromatography-mass spectrometry (UPLC-MS) experiments for metabolomics and lipidomics. Peaks that are likely to be associated with one compound can be clustered by the similarity of their chromatographic shape. Changes of concentration between sample groups can be inferred more accurately when multiple peaks are available. CONCLUSIONS: When the sample-size is limited, the proposed multi-peak approach improves the accuracy at inferring covariate effects. An R implementation and data are available at http://research.ics.aalto.fi/mi/software/peakANOVA/.

We extend kernelized matrix factorization with a full-Bayesian treatment and with an ability to work with multiple side information sources expressed as different kernels. Kernels have been introduced to integrate side information about the rows and columns, which is necessary for making out-of-matrix predictions. We discuss specifically binary output matrices but extensions to realvalued matrices are straightforward. We extend the state of the art in two key aspects: (i) A full-conjugate probabilistic formulation of the kernelized matrix factorization enables an efficient variational approximation, whereas full-Bayesian treatments are not computationally feasible in the earlier approaches. (ii) Multiple side information sources are included, treated as different kernels in multiple kernel learning which additionally reveals which side sources are informative. We then show that the framework can also be used for supervised and semi-supervised multilabel classification and multi-output regression, by considering samples and outputs as the domains where matrix factorization operates. Our method outperforms alternatives in predicting drug-protein interactions on two data sets. On multilabel classification, our algorithm obtains the lowest Hamming losses on 10 out of 14 data sets compared to five state-of-the-art multilabel classification algorithms. We finally show that the proposed approach outperforms alternatives in multi-output regression experiments on a yeast cell cycle data set.

We introduce interactive intent modeling, where the user directs exploratory search by providing feedback for estimates of search intents. The estimated intents are visualized for interaction on an Intent Radar, a novel visual interface that organizes intents onto a radial layout where relevant intents are close to the center of the visualization and similar intents have similar angles. The user can give feedback on the visualized intents, from which the system learns and visualizes improved intent estimates. We systematically evaluated the effect of the interactive intent modeling in a mixed-method task-based information seeking setting with 30 users, where we compared two interface variants for interactive intent modeling, namely intent radar and a simpler list-based interface, to a conventional search system. The results show that interactive intent modeling significantly improves users' task performance and the quality of retrieved information.

As the growth of stored personal digital information, such as photographs and emails, is continuously increasing, new tools for browsing and searching are needed. We introduce an intelligent mobile information access tool for personal data. The data are presented in an adaptive timeline where the displayed items function as search cues. The novelty is that the visualization is dynamically changed to emphasize relevant items, which makes them easier to recognize and select. The relevance is inferred during usage of the system from user feedback. In a user study, the dynamic timeline-based interface on a mobile device was shown to require less effort than conventional textual search.

We extend kernelized matrix factorization with a fully Bayesian treatment and with an ability to work with multiple side information sources expressed as different kernels. Kernel functions have been introduced to matrix factorization to integrate side information about the rows and columns (e.g., objects and users in recommender systems), which is necessary for making out-of-matrix (i.e., cold start) predictions. We discuss specifically bipartite graph inference, where the output matrix is binary, but extensions to more general matrices are straightforward. We extend the state of the art in two key aspects: (i) A fully conjugate probabilistic formulation of the kernelized matrix factorization problem enables an efficient variational approximation, whereas fully Bayesian treatments are not computationally feasible in the earlier approaches. (ii) Multiple side information sources are included, treated as different kernels in multiple kernel learning that additionally reveals which side information sources are informative. Our method outperforms alternatives in predicting drug-protein interactions on two data sets. We then show that our framework can also be used for solving multilabel learning problems by considering samples and labels as the two domains where matrix factorization operates on. Our algorithm obtains the lowest Hamming loss values on 10 out of 14 multilabel classification data sets compared to five state-of-the-art multilabel learning algorithms.

Neighbor embedding (NE) methods have found their use in data visualization but are limited in big data analysis tasks due to their O(n^2) complexity for n data samples. We demonstrate that the obvious approach of subsampling produces inferior results and propose a generic approximated optimization technique that reduces the NE optimization cost to O(n log n). The technique is based on realizing that in visualization the embedding space is necessarily very low-dimensional (2D or 3D), and hence efficient approximations developed for n-body force calculations can be applied. In gradient-based NE algorithms the gradient for an individual point decomposes into “forces” exerted by the other points. The contributions of close-by points need to be computed individually but far-away points can be approximated by their “center of mass”, rapidly computable by applying a recursive decomposition of the visualization space into quadrants. The new algorithm brings a significant speed-up for medium-size data, and brings “big data” within reach of visualization.

Techniques for both exploratory and known item search tend to direct only to more specific subtopics or individual documents, as opposed to allowing directing the exploration of the information space. We present an interactive information retrieval system that combines Reinforcement Learning techniques along with a novel user interface design to allow active engagement of users in directing the search. Users can directly manipulate document features (keywords) to indicate their interests and Reinforcement Learning is used to model the user by allowing the system to trade off between exploration and exploitation. This gives users the opportunity to more effectively direct their search nearer, further and following a direction. A task-based user study conducted with 20 participants comparing our system to a traditional query-based baseline indicates that our system significantly improves the effectiveness of information retrieval by providing access to more relevant and novel information without having to spend more time acquiring the information.

We study the feasibility of the following idea: Could a system learn to use the user's natural eye movements to infer relevance of real-world objects, if the user produced a set of learning data by clicking a "relevance" button during a learning session? If the answer is yes, the combination of eye tracking and machine learning would give a basis of "natural" interaction with the system by normally looking around, which would be very useful in mobile proactive setups. We measured the eye movements of the users while they were exploring an artificial art gallery. They labeled the relevant paintings by clicking a button while looking at them. The results show that a Gaussian process classifier accompanied by a time series kernel on the eye movements within an object predicts whether that object is relevant with better accuracy than dwell-time thresholding and random guessing.

MOTIVATION: Genome-wide measurement of transcript levels is an ubiquitous tool in biomedical research. As experimental data continues to be deposited in public databases, it is becoming important to develop search engines that enable the retrieval of relevant studies given a query study. While retrieval systems based on meta-data already exist, data-driven approaches that retrieve studies based on similarities in the expression data itself have a greater potential of uncovering novel biological insights. RESULTS: We propose an information retrieval method based on differential expression. Our method deals with arbitrary experimental designs and performs competitively with alternative approaches, while making the search results interpretable in terms of differential expression patterns. We show that our model yields meaningful connections between biological conditions from different studies. Finally, we validate a previously unknown connection between malignant pleural mesothelioma and SIM2s suggested by our method, via real-time polymerase chain reaction in an independent set of mesothelioma samples. AVAILABILITY: Supplementary data and source code are available from http://www.ebi.ac.uk/fg/research/rex.

MOTIVATION: Large public repositories of gene expression measurements offer the opportunity to position a new experiment into the context of earlier studies. While previous methods rely on experimental annotation or global similarity of expression profiles across genes or gene sets, we compare experiments by measuring similarity based on an unsupervised, data-driven regulatory model around pre-specified genes of interest. Our experiment retrieval approach is novel in two conceptual respects: (i) targetable focus and interpretability: the analysis is targeted at regulatory relationships of genes that are relevant to the analyst or come from prior knowledge; (ii) regulatory model-based similarity measure: related experiments are retrieved based on the strength of inferred regulatory links between genes. RESULTS: We learn a model for the regulation of specific genes from a data repository and exploit it to construct a similarity metric for an information retrieval task. We use the Fisher kernel, a rigorous similarity measure that typically has been applied to use generative models in discriminative classifiers. Results on human and plant microarray collections indicate that our method is able to substantially improve the retrieval of related experiments against standard methods. Furthermore, it allows the user to interpret biological conditions in terms of changes in link activity patterns. Our study of the osmotic stress network for Arabidopsis thaliana shows that the method successfully identifies relevant relationships around given key genes. AVAILABILITY: The code (R) is available at http://research.ics.tkk.fi/mi/software.shtml. CONTACT: [email protected]; [email protected]; [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Detailed and systematic understanding of the biological effects of millions of available compounds on living cells is a significant challenge. As most compounds impact multiple targets and pathways, traditional methods for analyzing structure-function relationships are not comprehensive enough. Therefore more advanced integrative models are needed for predicting biological effects elicited by specific chemical features. As a step towards creating such computational links we developed a data-driven chemical systems biology approach to comprehensively study the relationship of 76 structural 3D-descriptors (VolSurf, chemical space) of 1159 drugs with the microarray gene expression responses (biological space) they elicited in three cancer cell lines. The analysis covering 11350 genes was based on data from the Connectivity Map. We decomposed the biological response profiles into components, each linked to a characteristic chemical descriptor profile. RESULTS: Integrated analysis of both the chemical and biological space was more informative than either dataset alone in predicting drug similarity as measured by shared protein targets. We identified ten major components that link distinct VolSurf chemical features across multiple compounds to specific cellular responses. For example, component 2 (hydrophobic properties) strongly linked to DNA damage response, while component 3 (hydrogen bonding) was associated with metabolic stress. Individual structural and biological features were often linked to one cell line only, such as leukemia cells (HL-60) specifically responding to cardiac glycosides. CONCLUSIONS: In summary, our approach identified several novel links between specific chemical structure properties and distinct biological responses in cells incubated with these drugs. Importantly, the analysis focused on chemical-biological properties that emerge across multiple drugs. The decoding of such systematic relationships is necessary to build better models of drug effects, including unanticipated types of molecular properties having strong biological effects.

Recent evidence from serum metabolomics indicates that specific metabolic disturbances precede β-cell autoimmunity in humans and can be used to identify those children who subsequently progress to type 1 diabetes. The mechanisms behind these disturbances are unknown. Here we show the specificity of the pre-autoimmune metabolic changes, as indicated by their conservation in a murine model of type 1 diabetes. We performed a study in non-obese prediabetic (NOD) mice which recapitulated the design of the human study and derived the metabolic states from longitudinal lipidomics data. We show that female NOD mice who later progress to autoimmune diabetes exhibit the same lipidomic pattern as prediabetic children. These metabolic changes are accompanied by enhanced glucose-stimulated insulin secretion, normoglycemia, upregulation of insulinotropic amino acids in islets, elevated plasma leptin and adiponectin, and diminished gut microbial diversity of the Clostridium leptum group. Together, the findings indicate that autoimmune diabetes is preceded by a state of increased metabolic demands on the islets resulting in elevated insulin secretion and suggest alternative metabolic related pathways as therapeutic targets to prevent diabetes.

Probe defects are a major source of noise in gene expression studies. While existing approaches detect noisy probes based on external information such as genomic alignments, we introduce and validate a targeted probabilistic method for analyzing probe reliability directly from expression data and independently of the noise source. This provides insights into the various sources of probe-level noise and gives tools to guide probe design.

As prototypes of data glasses having both data augmentation and gaze tracking capabilities are becoming available, it is now possible to develop proactive gaze-controlled user interfaces to display information about objects, people, and other entities in real-world setups. In order to decide which objects the augmented information should be about, and how saliently to augment, the system needs an estimate of the importance or relevance of the objects of the scene for the user at a given time. The estimates will be used to minimize distraction of the user, and for providing efficient spatial management of the augmented items. This work is a feasibility study on inferring the relevance of objects in dynamic scenes from gaze. We collected gaze data from subjects watching a video for a pre-defined task. The results show that a simple ordinal logistic regression model gives relevance rankings of scene objects with a promising accuracy.

High-dimensional data are often visualized by dimensionality reduction methods whose goals are not directly related to visualization. We use a recent formalization of visualization as information retrieval and apply that formalism to data with structured annotations: we analyze gene expression data with annotations from the Gene Ontology (GO). We show that using the GO information in visualization yields better retrieval with respect to known ontological relationships and allows discovery of data properties not explained by the ontology.

MOTIVATION: Analysis of variance (ANOVA)-type methods are the default tool for the analysis of data with multiple covariates. These tools have been generalized to the multivariate analysis of high-throughput biological datasets, where the main challenge is the problem of small sample size and high dimensionality. However, the existing multi-way analysis methods are not designed for the currently increasingly important experiments where data is obtained from multiple sources. Common examples of such settings include integrated analysis of metabolic and gene expression profiles, or metabolic profiles from several tissues in our case, in a controlled multi-way experimental setup where disease status, medical treatment, gender and time-series are usual covariates. RESULTS: We extend the applicability area of multivariate, multi-way ANOVA-type methods to multi-source cases by introducing a novel Bayesian model. The method is capable of finding covariate-related dependencies between the sources. It assumes the measurements consist of groups of similarly behaving variables, and estimates the multivariate covariate effects and their interaction effects for the discovered groups of variables. In particular, the method partitions the effects to those shared between the sources and to source-specific ones. The method is specifically designed for datasets with small sample sizes and high dimensionality. We apply the method to a lipidomics dataset from a lung cancer study with two-way experimental setup, where measurements from several tissues with mostly distinct lipids have been taken. The method is also directly applicable to gene expression and proteomics. AVAILABILITY: An R-implementation is available at http://www.cis.hut.fi/projects/mi/software/multiWayCCA/.

MOTIVATION: Cell-biological processes are regulated through a complex network of interactions between genes and their products. The processes, their activating conditions and the associated transcriptional responses are often unknown. Organism-wide modeling of network activation can reveal unique and shared mechanisms between tissues, and potentially as yet unknown processes. The same method can also be applied to cell-biological conditions in one or more tissues. RESULTS: We introduce a novel approach for organism-wide discovery and analysis of transcriptional responses in interaction networks. The method searches for local, connected regions in a network that exhibit coordinated transcriptional response in a subset of tissues. Known interactions between genes are used to limit the search space and to guide the analysis. Validation on a human pathway network reveals physiologically coherent responses, functional relatedness between tissues and coordinated, context-specific regulation of the genes. AVAILABILITY: Implementation is freely available in R and Matlab at http://www.cis.hut.fi/projects/mi/software/NetResponse

We introduce relevant subtask learning, a new learning problem which is a variant of multi-task learning. The goal is to build a classifier for a task-of-interest for which we have too few training samples. We additionally have “supplementary data” c

Many recent works have combined two machine learning topics, learning of supervised distance metrics and manifold embedding methods, into supervised nonlinear dimensionality reduction methods. We show that a combination of an early metric learning method and a recent unsupervised dimensionality reduction method empirically outperforms previous methods. In our method, the Riemannian distance metric measures local change of class distributions, and the dimensionality reduction method makes a rigorous tradeoff between precision and recall in retrieving similar data points based on the reduced-dimensional display. The resulting supervised visualizations are good for finding (sets of) similar data samples that have similar class distributions.

Canonical correlation analysis (CCA) can be used to find correlating projections of two datasets with co-occurring samples. Instead of correlation, we would typically want to find more general dependencies, measured by mutual information. Variants of CCA based on non-parametric estimation of mutual information have been proposed previously; they outperform traditional CCA for non-Gaussian data but require infeasible amounts of computation for already quite modest sample sizes. We introduce a novel variant that uses a semi parametric estimate leading to a considerably faster algorithm. We apply the method on searching for statistical dependencies between multi-sensory stimuli and functional magnetic resonance imaging (fMRI) of brain activity- in contrast to using regression on either of them.

Several data analysis tools such as (kernel) canonical correlation analysis and various multi-view learning methods require paired observations in two data sets. We study the problem of inferring such pairing for data sets with no known one-to-one pairing. The pairing is found by an iterative algorithm that alternates between searching for feature representations that reveal statistical dependencies between the data sets, and finding the best pairs for the samples. The method is applied on pairing probe sets of two different microarray platforms.

We introduce GaZIR, a gaze-based interface for browsing and searching for images. The system computes on-line predictions of relevance of images based on implicit feedback, and when the user zooms in, the images predicted to be the most relevant are brought out. The key novelty is that the relevance feedback is inferred from implicit cues obtained in real-time from the gaze pattern, using an estimator learned during a separate training phase. The natural zooming interface can be connected to any content-based information retrieval engine operating on user feedback. We show with experiments on one engine that there is sufficient amount of information in the gaze patterns to make the estimated relevance feedback a viable choice to complement or even replace explicit feedback by pointing-and-clicking.

MOTIVATION: As ArrayExpress and other repositories of genome-wide experiments are reaching a mature size, it is becoming more meaningful to search for related experiments, given a particular study. We introduce methods that allow for the search to be based upon measurement data, instead of the more customary annotation data. The goal is to retrieve experiments in which the same biological processes are activated. This can be due either to experiments targeting the same biological question, or to as yet unknown relationships. RESULTS: We use a combination of existing and new probabilistic machine learning techniques to extract information about the biological processes differentially activated in each experiment, to retrieve earlier experiments where the same processes are activated and to visualize and interpret the retrieval results. Case studies on a subset of ArrayExpress show that, with a sufficient amount of data, our method indeed finds experiments relevant to particular biological questions. Results can be interpreted in terms of biological processes using the visualization techniques. AVAILABILITY: The code is available from http://www.cis.hut.fi/projects/mi/software/ismb09.

In the absence of explicit queries, an alternative is to try to infer users' interests from implicit feedback signals, such as clickstreams or eye tracking. The interests, formulated as an implicit query, can then be used in further searches. We formulate this task as a probabilistic model, which can be interpreted as a kind of transfer or meta-learning. The probabilistic model is demonstrated to outperform an earlier kernel-based method in a small-scale information retrieval task.

BACKGROUND: Bioinformatics data analysis toolbox needs general-purpose, fast and easily interpretable preprocessing tools that perform data integration during exploratory data analysis. Our focus is on vector-valued data sources, each consisting of measurements of the same entity but on different variables, and on tasks where source-specific variation is considered noisy or not interesting. Principal components analysis of all sources combined together is an obvious choice if it is not important to distinguish between data source-specific and shared variation. Canonical Correlation Analysis (CCA) focuses on mutual dependencies and discards source-specific "noise" but it produces a separate set of components for each source. RESULTS: It turns out that components given by CCA can be combined easily to produce a linear and hence fast and easily interpretable feature extraction method. The method fuses together several sources, such that the properties they share are preserved. Source-specific variation is discarded as uninteresting. We give the details and implement them in a software tool. The method is demonstrated on gene expression measurements in three case studies: classification of cell cycle regulated genes in yeast, identification of differentially expressed genes in leukemia, and defining stress response in yeast. The software package is available at http://www.cis.hut.fi/projects/mi/software/drCCA/. CONCLUSION: We introduced a method for the task of data fusion for exploratory data analysis, when statistical dependencies between the sources and not within a source are interesting. The method uses canonical correlation analysis in a new way for dimensionality reduction, and inherits its good properties of being simple, fast, and easily interpretable as a linear projection.

We introduce a mixture of probabilistic canonical correlation analyzers model for analyzing local correlations, or more generally mutual statistical dependencies, in cooccurring data pairs. The model extends the traditional canonical correlation analysis and its probabilistic interpretation in three main ways. First, a full Bayesian treatment enables analysis of small samples (large p, small n, a crucial problem in bioinformatics, for instance), and rigorous estimation of the degree of dependency and independency. Secondly, the mixture formulation generalizes the method from global linearity to the more reasonable assumption of different kinds of dependencies for different kinds of data. As a third novel extension the method decomposes the variation in the data into shared and data set-specific components. 1.

This supplement contains extended versions of a selected subset of papers presented at the workshop PMSB 2007, Probabilistic Modeling and Machine Learning in Structural and Systems Biology, Tuusula, Finland, from June 17 to 18, 2006.

BACKGROUND: Human endogenous retroviruses (HERVs) are surviving traces of ancient retrovirus infections and now reside within the human DNA. Recently HERV expression has been detected in both normal tissues and diseased patients. However, the activities (expression levels) of individual HERV sequences are mostly unknown. RESULTS: We introduce a generative mixture model, based on Hidden Markov Models, for estimating the activities of the individual HERV sequences from EST (expressed sequence tag) databases. We use the model to estimate the relative activities of 181 HERVs. We also empirically justify a faster heuristic method for HERV activity estimation and use it to estimate the activities of 2450 HERVs. The majority of the HERV activities were previously unknown. CONCLUSION: (i) Our methods estimate activity accurately based on experiments on simulated data. (ii) Our estimate on real data shows that 7% of the HERVs are active. The active ones are spread unevenly into HERV groups and relatively uniformly in terms of estimated age. HERVs with the retroviral env gene are more often active than HERVs without env. Few of the active HERVs have open reading frames for retroviral proteins.

Canonical correlation analysis (CCA) is equivalent to finding mutual information-maximizing projections for normally distributed data. We remove the restriction of normality by non-parametric estimation, and formulate the problem of finding dependent components with a connection to Bayes factors. The method is applied for characterizing yeast stress by finding what is in common in several different stress conditions.

We introduce an expectation maximizationtype (EM) algorithm for maximum likelihood optimization of conditional densities. It is applicable to hidden variable models where the distributions are from the exponential family. The algorithm can alternatively be viewed as automatic step size selection for gradient ascent, where the amount of computation is traded off to guarantees that each step increases the likelihood. The tradeoff makes the algorithm computationally more feasible than the earlier conditional EM. The method gives a theoretical basis for extended Baum Welch algorithms used in discriminative hidden Markov models in speech recognition, and compares favourably with the current best method in the experiments.

We study a new task, proactive information retrieval by combining implicit relevance feedback and collaborative filtering. We have constructed a controlled experimental setting, a prototype application, in which the users try to find interesting scientific articles by browsing their titles. Implicit feedback is inferred from eye movement signals, with discriminative hidden Markov models estimated from existing data in which explicit relevance feedback is available. Collaborative filtering is carried out using the User Rating Profile model, a state-of-the-art probabilistic latent variable model, computed using Markov Chain Monte Carlo techniques. For new document titles the prediction accuracy with eye movements, collaborative filtering, and their combination was significantly better than by chance. The best prediction accuracy still leaves room for improvement but shows that proactive information retrieval and combination of many sources of relevance feedback is feasible.

We model dependencies between m multivariate continuous-valued information sources by a combination of (i) a generalized canonical correlations analysis (gCCA) to reduce dimensionality while preserving dependencies in m - 1 of them, and (ii) summarizing dependencies with the remaining one by associative clustering. This new combination of methods avoids multiway associative clustering which would require a multiway contingency table and hence suffer from curse of dimensionality of the table. The method is applied to summarizing properties of yeast stress by searching for dependencies (commonalities) between expression of genes of baker's yeast Saccharomyces cerevisiae in various stressful treatments, and summarizing stress regulation by finally adding data about transcription factor binding sites.

About 8 per cent of the human genome consists of human endogenous retroviral sequences (HERVs), which are remains from ancient infections. The HERVs may give rise to transcripts or affect the expression of human genes. The first step in understanding HERV function is to classify HERVs into families. In this work we study the relationships of existing HERV families and detect potentially new HERV families. A Median Self-Organizing Map (SOM), a SOM for non-vectorial data, is used to group and visualize a collection of 3661 HERVs. The SOM-based analysis is complemented with estimates of the reliability of the results. A novel trustworthiness visualization method is used to estimate which parts of the SOM visualization are reliable and which not. The reliability of extracted interesting HERV groups is verified by a bootstrap procedure suitable for SOM visualization-based analysis. The SOM detects a group of epsilonretroviral sequences and a group of ERV9, HERVW, and HUERSP3 sequences which suggests that ERV9 and HERVW sequences may have a common origin.

High-throughput genomic measurements, interpreted as cooccurring data samples from multiple sources, open up a fresh problem for machine learning: What is in common in the different data sets, that is, what kind of statistical dependencies are there between the paired samples from the different sets? We introduce a clustering algorithm for exploring the dependencies. Samples within each data set are grouped such that the dependencies between groups of different sets capture as much of pairwise dependencies between the samples as possible. We formalize this problem in a novel probabilistic way, as optimization of a Bayes factor. The method is applied to reveal commonalities and exceptions in gene expression between organisms and to suggest regulatory interactions in the form of dependencies between gene expression profiles and regulator binding patterns.

A simple probabilistic model is introduced to generalize classical linear discriminant analysis (LDA) in finding components that are informative of or relevant for data classes. The components maximize the predictability of the class distribution which is asymptotically equivalent to 1) maximizing mutual information with the classes, and 2) finding principal components in the so-called learning or Fisher metrics. The Fisher metric measures only distances that are relevant to the classes, that is, distances that cause changes in the class distribution. The components have applications in data exploration, visualization, and dimensionality reduction. In empirical experiments, the method outperformed, in addition to more classical methods, a Renyi entropy-based alternative while having essentially equivalent computational cost.

About eight percent of the human genome consists of human endogenous retrovirus sequences. Human endogenous retroviruses (HERV) are remains from ancient infections by retroviruses. The HERVs are mutated and deficient, but they still may give rise to transcripts or may affect the expression of human genes. The HERVs stem from several kinds of retroviruses., The possible current functioning of the HERV sequences may reflect the origin of the HERVs. Hence, the classification of the diverse HERV sequences is a natural starting point when investigating the effect of HERVs in humans. The current HERV taxonomy is incomplete: some sequences cannot be assigned to any class and the classification is ambiguous for others. A median self-organizing map (SOM), a SOM for data about pairwise distances between samples, can be used to group all the HERVs found in the human genome. It visualizes the collection of 3661 HERV sequences found by the RetroTector system, on a two-dimensional display that represents similarity relationships between individual sequences, as well as cluster structures and similarities of clusters. The SOM, as any dimensionality reduction method, necessarily has to make compromises when representing the data. In this work we extend the visualizations by bootstrap-based estimates on which parts of the visualization are reliable and which not, and use the SOM to find potentially new HERV groups.

The sequential information bottleneck (sIB) algorithm clusters co-occurrence data such as text documents vs. words. We introduce a variant that models sparse co-occurrence data by a generative process. This turns the objective function of sIB, mutual information, into a Bayes factor, while keeping it intact asymptotically, for non-sparse data. Experimental performance of the new algorithm is comparable to the original sIB for large data sets, and better for smaller, sparse sets.

This work is part of a proactive information retrieval project that aims at estimating relevance from implicit user feedback. The noisy feedback signal needs to be complemented with all available information, and textual content is one of the natural sources. Here we take the first steps by investigating whether this source is at all useful in the challenging setting of estimating the relevance of a new document based on only few samples with known relevance. It turns out that even sophisticated unsupervised methods like multinomial PCA (or latent Dirichlet allocation) cannot help much. By contrast, feature extraction supervised by relevant auxiliary data may help.

BACKGROUND: Conventionally, the first step in analyzing the large and high-dimensional data sets measured by microarrays is visual exploration. Dendrograms of hierarchical clustering, self-organizing maps (SOMs), and multidimensional scaling have been used to visualize similarity relationships of data samples. We address two central properties of the methods: (i) Are the visualizations trustworthy, i.e., if two samples are visualized to be similar, are they really similar? (ii) The metric. The measure of similarity determines the result; we propose using a new learning metrics principle to derive a metric from interrelationships among data sets. RESULTS: The trustworthiness of hierarchical clustering, multidimensional scaling, and the self-organizing map were compared in visualizing similarity relationships among gene expression profiles. The self-organizing map was the best except that hierarchical clustering was the most trustworthy for the most similar profiles. Trustworthiness can be further increased by treating separately those genes for which the visualization is least trustworthy. We then proceed to improve the metric. The distance measure between the expression profiles is adjusted to measure differences relevant to functional classes of the genes. The genes for which the new metric is the most different from the usual correlation metric are listed and visualized with one of the visualization methods, the self-organizing map, computed in the new metric. CONCLUSIONS: The conjecture from the methodological results is that the self-organizing map can be recommended to complement the usual hierarchical clustering for visualizing and exploring gene expression data. Discarding the least trustworthy samples and improving the metric still improves it.

We study the problem of learning groups or categories that are local in the continuous primary space but homogeneous by the distributions of an associated auxiliary random variable over a discrete auxiliary space. Assuming that variation in the auxiliary space is meaningful, categories will emphasize similarly meaningful aspects of the primary space. From a data set consisting of pairs of primary and auxiliary items, the categories are learned by minimizing a Kullback-Leibler divergence-based distortion between (implicitly estimated) distributions of the auxiliary data, conditioned on the primary data. Still, the categories are defined in terms of the primary space. An online algorithm resembling the traditional Hebb-type competitive learning is introduced for learning the categories. Minimizing the distortion criterion turns out to be equivalent to maximizing the mutual information between the categories and the auxiliary data. In addition, connections to density estimation and to the distributional clustering paradigm are outlined. The method is demonstrated by clustering yeast gene expression data from DNA chips, with biological knowledge about the functional classes of the genes as the auxiliary data.

We introduce a method for deriving a metric, locally based on the Fisher information matrix, into the data space. A self-organizing map (SOM) is computed in the new metric to explore financial statements of enterprises. The metric measures local distances in terms of changes in the distribution of an auxiliary random variable that reflects what is important in the data. In this paper the variable indicates bankruptcy within the next few years. The conditional density of the auxiliary variable is first estimated, and the change in the estimate resulting from local displacements in the primary data space is measured using the Fisher information matrix. When a self-organizing map is computed in the new metric it still visualizes the data space in a topology-preserving fashion, but represents the (local) directions in which the probability of bankruptcy changes the most.

We introduce an algorithm for learning a local metric to a continuous input space that measures distances in terms of relevance to the processing task. The relevance is defined as local changes in discrete auxiliary information, which may be for example the class of the data items, an index of performance, or a contextual input. A set of neurons first learns representations that maximize the mutual information between their outputs and the random variable representing the auxiliary information. The implicit knowledge gained about relevance is then transformed into a new metric of the input space that measures the change in the auxiliary information in the sense of local approximations to the Kullback-Leibler divergence. The new metric can be used in further processing by other algorithms. It is especially useful in data analysis applications since the distances can be interpreted in terms of the local relevance of the original variables.

This article describes the implementation of a system that is able to organize vast document collections according to textual similarities. It is based on the self-organizing map (SOM) algorithm. As the feature vectors for the documents statistical representations of their vocabularies are used. The main goal in our work has been to scale up the SOM algorithm to be able to deal with large amounts of high-dimensional data. In a practical experiment we mapped 6,840,568 patent abstracts onto a 1,002,240-node SOM. As the feature vectors we used 500-dimensional vectors of stochastic figures obtained as random projections of weighted word histograms.

The adaptive-subspace self-organizing map (ASSOM) is a modular neural network architecture, the modules of which learn to identify input patterns subject to some simple transformations. The learning process is unsupervised, competitive, and related to that of the traditional SOM (self-organizing map). Each neural module becomes adaptively specific to some restricted class of transformations, and modules close to each other in the network become tuned to similar features in an orderly fashion. If different transformations exist in the input signals, different subsets of ASSOM units become tuned to these transformation classes.

Vector quantization (VQ) can be used to compress images with high compression ratios. The VQ methods produce a sequence of code values which identifies the codebook model vectors to be used as blocks of pixels in the decoded image. In this paper we define a novel non-lossy and computationally efficient method to further compress the code sequence based on the relative frequencies of the code values, and the spatial distribution of each code. In an example case we reduced the bit rate by 29%. A further reduction of 7 percentage units was obtained when the VQ codebook was produced by the self-organizing map (SOM) algorithm. A SOM codebook has the property that similar blocks have similar codes, which was used to take advantage of spatial redundancies in the image.

We have implemented a novel combination of LVQ--codebooks and hidden Markov models (HMMs) for transcribing spoken Finnish words into phonemes. A separate LVQ--codebook was trained for the unvoiced plosives /k/, /p/ and /t/. HMMs were then modified to accommodate two parallel phoneme codebooks and a third codebook representing the power of the speech signal. The error rate of phonemic transcription was as low as 9.1 %. A preliminary test with an additional /b/, /d/ and /g/ codebook was also carried out. 1 INTRODUCTION LVQ: Ordinary vector quantizers replace an input vector with a discrete symbol corresponding to the closest codebook vector. Similar minimum--distance rule can also be used in pattern classification, if the symbols are made to represent different classes of data. Several codebook vectors may then belong to the same class. Although vector quantizers and minimum--distance classifiers are closely related, their design criteria are quite different. Learning Vector Quantizat...