We develop VSD, a method for conditioning a generative model of discrete, combinatorial designs on a rare desired class by efficiently evaluating a black-box (e.g. experiment, simulation) in a batch sequential manner. We call this task active generation; we formalize active generation's requirements and desiderata, and formulate a solution via variational inference. VSD uses off-the-shelf gradient based optimization routines, can learn powerful generative models for desirable designs, and can take advantage of scalable predictive models. We derive asymptotic convergence rates for learning the true conditional generative distribution of designs with certain configurations of our method. After illustrating the generative model on images, we empirically demonstrate that VSD can outperform existing baseline methods on a set of real sequence-design problems in various protein and DNA/RNA engineering tasks.

We introduce active generation of Pareto sets (A-GPS), a new framework for online discrete black-box multi-objective optimization (MOO). A-GPS learns a generative model of the Pareto set that supports a-posteriori conditioning on user preferences. The method employs a class probability estimator (CPE) to predict non-dominance relations and to condition the generative model toward high-performing regions of the search space. We also show that this non-dominance CPE implicitly estimates the probability of hypervolume improvement (PHVI). To incorporate subjective trade-offs, A-GPS introduces _preference direction vectors_ that encode user-specified preferences in objective space. At each iteration, the model is updated using both Pareto membership and alignment with these preference directions, producing an amortized generative model capable of sampling across the Pareto front without retraining. The result is a simple yet powerful approach that achieves high-quality Pareto set approximations, avoids explicit hypervolume computation, and flexibly captures user preferences. Empirical results on synthetic benchmarks and protein design tasks demonstrate strong sample efficiency and effective preference incorporation.

Squared families of probability distributions have been studied and applied in numerous machine learning contexts. Typically, they appear as likelihoods, where their advantageous computational, geometric and statistical properties are exploited for fast estimation algorithms, representational properties and statistical guarantees. Here, we investigate the use of squared families as prior beliefs in Bayesian inference. We find that they can form helpful conjugate families, often allowing for closed-form and tractable Bayesian inference and marginal likelihoods. We apply such conjugate families to Bayesian regression in feature space using end-to-end learnable neural network features. Such a setting allows for a rich multi-modal alternative to Gaussian processes with neural network features, often called deep kernel learning. We demonstrate our method on few shot learning, outperforming existing neural methods based on Gaussian processes and normalising flows.

We introduce squared neural Poisson point processes (SNEPPPs) by parameterising the intensity function by the squared norm of a two layer neural network. When the hidden layer is fixed and the second layer has a single neuron, our approach resembles previous uses of squared Gaussian process or kernel methods, but allowing the hidden layer to be learnt allows for additional flexibility. In many cases of interest, the integrated intensity function admits a closed form and can be computed in quadratic time in the number of hidden neurons. We enumerate a far more extensive number of such cases than has previously been discussed. Our approach is more memory and time efficient than naive implementations of squared or exponentiated kernel methods or Gaussian processes. Maximum likelihood and maximum a posteriori estimates in a reparameterisation of the final layer of the intensity function can be obtained by solving a (strongly) convex optimisation problem using projected gradient descent. We demonstrate SNEPPPs on real, and synthetic benchmarks, and provide a software implementation.

Principal Component Analysis (PCA) and its exponential family extensions have three components: observations, latents and parameters of a linear transformation. We consider a generalised setting where the canonical parameters of the exponential family are a nonlinear transformation of the latents. We show explicit relationships between particular neural network architectures and the corresponding statistical models. We find that deep equilibrium models — a recently introduced class of implicit neural networks — solve maximum a-posteriori (MAP) estimates for the latents and parameters of the transformation. Our analysis provides a systematic way to relate activation functions, dropout, and layer structure, to statistical assumptions about the observations, thus providing foundational principles for unsupervised DEQs. For hierarchical latents, individual neurons can be interpreted as nodes in a deep graphical model. Our DEQ feature maps are end-to-end differentiable, enabling fine-tuning for downstream tasks.

Flexible models for probability distributions are an essential ingredient in many machine learning tasks. We develop and investigate a new class of probability distributions, which we call a Squared Neural Family (SNEFY), formed by squaring the 2-norm of a neural network and normalising it with respect to a base measure. Following the reasoning similar to the well established connections between infinitely wide neural networks and Gaussian processes, we show that SNEFYs admit closed form normalising constants in many cases of interest, thereby resulting in flexible yet fully tractable density models. SNEFYs strictly generalise classical exponential families, are closed under conditioning, and have tractable marginal distributions. Their utility is illustrated on a variety of density estimation, conditional density estimation, and density estimation with missing data tasks.

We consider the continuum-armed bandits problem, under a novel setting of recommending the best arms within a fixed budget under aggregated feedback. This is motivated by applications where the precise rewards are impossible or expensive to obtain, while an aggregated reward or feedback, such as the average over a subset, is available. We constrain the set of reward functions by assuming that they are from a Gaussian Process and propose the Gaussian Process Optimistic Optimisation (GPOO) algorithm. We adaptively construct a tree with nodes as subsets of the arm space, where the feedback is the aggregated reward of representatives of a node. We propose a new simple regret notion with respect to aggregated feedback on the recommended arms. We provide theoretical analysis for the proposed algorithm, and recover single point feedback as a special case. We illustrate GPOO and compare it with related algorithms on simulated data.

We consider a variant of the best arm identification task in stochastic multi-armed bandits. Motivated by risk-averse decision-making problems, our goal is to identify a set of $m$ arms with the highest $\tau$-quantile values within a fixed budget. We prove asymmetric two-sided concentration inequalities for order statistics and quantiles of random variables that have non-decreasing hazard rate, which may be of independent interest. With these inequalities, we analyse a quantile version of Successive Accepts and Rejects (Q-SAR). We derive an upper bound for the probability of arm misidentification, the first justification of a quantile based algorithm for fixed budget multiple best arms identification. We show illustrative experiments for best arm identification.

We propose a new model for networks of time series that influence each other. Graph structures among time series are found in diverse domains, such as web traffic influenced by hyperlinks, product sales influenced by recommendation, or urban transport volume influenced by road networks and weather. There has been recent progress in graph modeling and in time series forecasting, respectively, but an expressive and scalable approach for a network of series does not yet exist. We introduce Radflow, a novel model that embodies three key ideas: a recurrent neural network to obtain node embeddings that depend on time, the aggregation of the flow of influence from neighboring nodes with multi-head attention, and the multi-layer decomposition of time series. Radflow naturally takes into account dynamic networks where nodes and edges change over time, and it can be used for prediction and data imputation tasks. On real-world datasets ranging from a few hundred to a few hundred thousand nodes, we observe that Radflow variants are the best performing model across a wide range of settings. The recurrent component in Radflow also outperforms N-BEATS, the state-of-the-art time series model. We show that Radflow can learn different trends and seasonal patterns, that it is robust to missing nodes and edges, and that correlated temporal patterns among network neighbors reflect influence strength. We curate WikiTraffic, the largest dynamic network of time series with 366K nodes and 22M time-dependent links spanning five years. This dataset provides an open benchmark for developing models in this area, with applications that include optimizing resources for the web. More broadly, Radflow has the potential to improve forecasts in correlated time series networks such as the stock market, and impute missing measurements in geographically dispersed networks of natural phenomena.

MOTIVATION: The high accuracy of recent haplotype phasing tools is enabling the integration of haplotype (or phase) information more widely in genetic investigations. One such possibility is phase-aware expression quantitative trait loci (eQTL) analysis, where haplotype-based analysis has the potential to detect associations that may otherwise be missed by standard SNP-based approaches. RESULTS: We present eQTLHap, a novel method to investigate associations between gene expression and genetic variants, considering their haplotypic and genotypic effect. Using multiple simulations based on real data, we demonstrate that phase-aware eQTL analysis significantly outperforms typical SNP-based methods when the causal genetic architecture involves multiple SNPs. We show that phase-aware eQTL analysis is robust to phasing errors, showing only a minor impact ($<4\%$) on sensitivity. Applying eQTLHap to real GEUVADIS and GTEx datasets detects numerous novel eQTLs undetected by a single-SNP approach, with 22 eQTLs replicating across studies or tissue types, highlighting the utility of phase-aware eQTL analysis. AVAILABILITY AND IMPLEMENTATION: https://github.com/ziadbkh/eQTLHap. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Briefings in Bioinformatics online.

Machine learning algorithms are increasingly used to make or support important decisions about people's lives. This has led to interest in the problem of fair classification, which involves learning to make decisions that are non-discriminatory with respect to a sensitive variable such as race or gender. Several methods have been proposed to solve this problem, including fair representation learning, which cleans the input data used by the algorithm to remove information about the sensitive variable. We show that using fair representation learning as an intermediate step in fair classification incurs a cost compared to directly solving the problem, which we refer to as thecost of mistrust. We show that fair representation learning in fact addresses a different problem, which is of interest when the data user is not trusted to access the sensitive variable. We quantify the benefits of fair representation learning, by showing that any subsequent use of the cleaned data will not be too unfair. The benefits we identify result from restricting the decisions of adversarial data users, while the costs are due to applying those same restrictions to other data users.

The last few years have seen a staggering number of empirical studies of the robustness of neural networks in a model of adversarial perturbations of their inputs. Most rely on an adversary which carries out local modifications within prescribed balls. None however has so far questioned the broader picture: how to frame a resource-bounded adversary so that it can be severely detrimental to learning, a non-trivial problem which entails at a minimum the choice of loss and classifiers. We suggest a formal answer for losses that satisfy the minimal statistical requirement of being proper. We pin down a simple sufficient property for any given class of adversaries to be detrimental to learning, involving a central measure of “harmfulness” which generalizes the well-known class of integral probability metrics. A key feature of our result is that it holds for all proper losses, and for a popular subset of these, the optimisation of this central measure appears to be independent of the loss. When classifiers are Lipschitz – a now popular approach in adversarial training –, this optimisation resorts to optimal transport to make a low-budget compression of class marginals. Toy experiments reveal a finding recently separately observed: training against a sufficiently budgeted adversary of this kind improves generalization.

Individual characteristics in human decision-making are often quantified by fitting a parametric cognitive model to subjects' behavior and then studying differences between them in the associated parameter space. However, these models often fit behavior more poorly than recurrent neural networks (RNNs), which are more flexible and make fewer assumptions about the underlying decision-making processes. Unfortunately, the parameter and latent activity spaces of RNNs are generally high-dimensional and uninterpretable, making it hard to use them to study individual differences. Here, we show how to benefit from the flexibility of RNNs while representing individual differences in a low-dimensional and interpretable space. To achieve this, we propose a novel end-to-end learning framework in which an encoder is trained to map the behavior of subjects into a low-dimensional latent space. These low-dimensional representations are used to generate the parameters of individual RNNs corresponding to the decision-making process of each subject. We introduce terms into the loss function that ensure that the latent dimensions are informative and disentangled, i.e., encouraged to have distinct effects on behavior. This allows them to align with separate facets of individual differences. We illustrate the performance of our framework on synthetic data as well as a dataset including the behavior of patients with psychiatric disorders.

We consider the problem of learning a low dimensional representation for compositional data. Compositional data consists of a collection of nonnegative data that sum to a constant value. Since the parts of the collection are statistically dependent, many standard tools cannot be directly applied. Instead, compositional data must be first transformed before analysis. Focusing on principal component analysis (PCA), we propose an approach that allows low dimensional representation learning directly from the original data. Our approach combines the benefits of the log-ratio transformation from compositional data analysis and exponential family PCA. A key tool in its derivation is a generalization of the scaled Bregman theorem, that relates the perspective transform of a Bregman divergence to the Bregman divergence of a perspective transform and a remainder conformal divergence. Our proposed approach includes a convenient surrogate (upper bound) loss of the exponential family PCA which has an easy to optimize form. We also derive the corresponding form for nonlinear autoencoders. Experiments on simulated data and microbiome data show the promise of our method.

Wheel defects on railway wagons have been identified as an important source of damage to the railway infrastructure and rolling stock. They also cause noise and vibration emissions that are costly to mitigate. We propose two machine learning methods to automatically detect these wheel defects, based on the wheel vertical force measured by a permanently installed sensor system on the railway network. Our methods automatically learn different types of wheel defects and predict during normal operation if a wheel has a defect or not. The first method is based on novel features for classifying time series data and it is used for classification with a support vector machine. To evaluate the performance of our method we construct multiple data sets for the following defect types: flat spot, shelling, and non-roundness. We outperform classical defect detection methods for flat spots and demonstrate prediction for the other two defect types for the first time. Motivated by the recent success of artificial neural networks for image classification, we train custom artificial neural networks with convolutional layers on 2-D representations of the measurement time series. The neural network approach improves the performance on wheels with flat spots and non-roundness by explicitly modeling the multi sensor structure of the measurement system through multiple instance learning and shift invariant networks.

We present an interactive visualisation tool for recommending travel trajectories. This system is based on new machine learning formulations and algorithms for the sequence recommendation problem. The system starts from a map-based overview, taking an interactive query as starting point. It then breaks down contributions from different geographical and user behavior features, and those from individual points-of-interest versus pairs of consecutive points on a route. The system also supports detailed quantitative interrogation by comparing a large number of features for multiple points. Effective trajectory visualisations can potentially benefit a large cohort of online map users and assist their decision-making. More broadly, the design of this system can inform visualisations of other structured prediction tasks, such as for sequences or trees.

Modern genomics techniques generate overwhelming quantities of data. Extracting population genetic variation demands computationally efficient methods to determine genetic relatedness between individuals (or "samples") in an unbiased manner, preferably de novo. Rapid estimation of genetic relatedness directly from sequencing data has the potential to overcome reference genome bias, and to verify that individuals belong to the correct genetic lineage before conclusions are drawn using mislabelled, or misidentified samples. We present the k-mer Weighted Inner Product (kWIP), an assembly-, and alignment-free estimator of genetic similarity. kWIP combines a probabilistic data structure with a novel metric, the weighted inner product (WIP), to efficiently calculate pairwise similarity between sequencing runs from their k-mer counts. It produces a distance matrix, which can then be further analysed and visualised. Our method does not require prior knowledge of the underlying genomes and applications include establishing sample identity and detecting mix-up, non-obvious genomic variation, and population structure. We show that kWIP can reconstruct the true relatedness between samples from simulated populations. By re-analysing several published datasets we show that our results are consistent with marker-based analyses. kWIP is written in C++, licensed under the GNU GPL, and is available from https://github.com/kdmurray91/kwip.

The problem of recommending tours to travellers is an important and broadly studied area. Suggested solutions include various approaches of points-of-interest (POI) recommendation and route planning. We consider the task of recommending a sequence of POIs, that simultaneously uses information about POIs and routes. Our approach unifies the treatment of various sources of information by representing them as features in machine learning algorithms, enabling us to learn from past behaviour. Information about POIs are used to learn a POI ranking model that accounts for the start and end points of tours. Data about previous trajectories are used for learning transition patterns between POIs that enable us to recommend probable routes. In addition, a probabilistic model is proposed to combine the results of POI ranking and the POI to POI transitions. We propose a new F1 score on pairs of POIs that capture the order of visits. Empirical results show that our approach improves on recent methods, and demonstrate that combining points and routes enables better trajectory recommendations.

Knowledge graph construction consists of two tasks: extracting information from external resources (knowledge population) and inferring missing information through a statistical analysis on the extracted information (knowledge completion). In many cases, insufficient external resources in the knowledge population hinder the subsequent statistical inference. The gap between these two processes can be reduced by an incremental population approach. We propose a new probabilistic knowledge graph factorisation method that benefits from the path structure of existing knowledge (e.g. syllogism) and enables a common modelling approach to be used for both incremental population and knowledge completion tasks. More specifically, the probabilistic formulation allows us to develop an incremental population algorithm that trades off exploitation-exploration. Experiments on three benchmark datasets show that the balanced exploitation-exploration helps the incremental population, and the additional path structure helps to predict missing information in knowledge completion.

We propose an extension to Hawkes processes by treating the levels of self-excitation as a stochastic differential equation. Our new point process allows better approximation in application domains where events and intensities accelerate each other with correlated levels of contagion. We generalize a recent algorithm for simulating draws from Hawkes processes whose levels of excitation are stochastic processes, and propose a hybrid Markov chain Monte Carlo approach for model fitting. Our sampling procedure scales linearly with the number of required events and does not require stationarity of the point process. A modular inference procedure consisting of a combination between Gibbs and Metropolis Hastings steps is put forward. We recover expectation maximization as a special case. Our general approach is illustrated for contagion following geometric Brownian motion and exponential Langevin dynamics.

Given samples from two densities p and q, density ratio estimation (DRE) is the problem of estimating the ratio p/q. Two popular discriminative approaches to DRE are KL importance estimation (KLIEP), and least squares importance fitting (LSIF). In this paper, we show that KLIEP and LSIF both employ class-probability estimation (CPE) losses. Motivated by this, we formally relate DRE and CPE, and demonstrate the viability of using existing losses from one problem for the other. For the DRE problem, we show that essentially any CPE loss (eg logistic, exponential) can be used, as this equivalently minimises a Bregman divergence to the true density ratio. We show how different losses focus on accurately modelling different ranges of the density ratio, and use this to design new CPE losses for DRE. For the CPE problem, we argue that the LSIF loss is useful in the regime where one wishes to rank instances with maximal accuracy at the head of the ranking. In the course of our analysis, we establish a Bregman divergence identity that may be of independent interest.

Bregman divergences play a central role in the design and analysis of a range of machine learning algorithms through a handful of popular theorems. We present a new theorem which shows that ``Bregman distortions'' (employing a potentially non-convex generator) may be exactly re-written as a scaled Bregman divergence computed over transformed data. This property can be viewed from the standpoints of geometry (a scaled isometry with adaptive metrics) or convex optimization (relating generalized perspective transforms). Admissible distortions include {geodesic distances} on curved manifolds and projections or gauge-normalisation. Our theorem allows one to leverage to the wealth and convenience of Bregman divergences when analysing algorithms relying on the aforementioned Bregman distortions. We illustrate this with three novel applications of our theorem: a reduction from multi-class density ratio to class-probability estimation, a new adaptive projection free yet norm-enforcing dual norm mirror descent algorithm, and a reduction from clustering on flat manifolds to clustering on curved manifolds. Experiments on each of these domains validate the analyses and suggest that the scaled Bregman theorem might be a worthy addition to the popular handful of Bregman divergence properties that have been pervasive in machine learning.

MOTIVATION: Understanding the occurrence and regulation of alternative splicing (AS) is a key task towards explaining the regulatory processes that shape the complex transcriptomes of higher eukaryotes. With the advent of high-throughput sequencing of RNA (RNA-Seq), the diversity of AS transcripts could be measured at an unprecedented depth. Although the catalog of known AS events has grown ever since, novel transcripts are commonly observed when working with less well annotated organisms, in the context of disease, or within large populations. Whereas an identification of complete transcripts is technically challenging and computationally expensive, focusing on single splicing events as a proxy for transcriptome characteristics is fruitful and sufficient for a wide range of analyses. RESULTS: We present SplAdder, an alternative splicing toolbox, that takes RNA-Seq alignments and an annotation file as input to (i) augment the annotation based on RNA-Seq evidence, (ii) identify alternative splicing events present in the augmented annotation graph, (iii) quantify and confirm these events based on the RNA-Seq data and (iv) test for significant quantitative differences between samples. Thereby, our main focus lies on performance, accuracy and usability. AVAILABILITY: Source code and documentation are available for download at http://github.com/ratschlab/spladder Example data, introductory information and a small tutorial are accessible via http://bioweb.me/spladder CONTACTS: : [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

We study the problem of rank aggregation: given a set of ranked lists, we want to form a consensus ranking. Furthermore, we consider the case of extreme lists: i.e., only the rank of the best or worst elements are known. We impute missing ranks and generalise Spearman's $\rho$ to extreme ranks. Our main contribution is the derivation of a non-parametric estimator for rank aggregation based on multivariate extensions of Spearman's $\rho$, which measures correlation between a set of ranked lists. Multivariate Spearman's $\rho$ is defined using copulas, and we show that the geometric mean of normalised ranks maximises multivariate correlation. Motivated by this, we propose a weighted geometric mean approach for learning to rank which has a closed form least squares solution. When only the best (top-k) or worst (bottom-k) elements of a ranked list are known, we impute the missing ranks by the average value, allowing us to apply Spearman's $\rho$. We discuss an optimistic and pessimistic imputation of missing values, which respectively maximise and minimise correlation, and show its effect on aggregating university rankings. Finally, we demonstrate good performance on the rank aggregation benchmarks MQ2007 and MQ2008.

This paper presents a theoretical foundation for an SVM solver in Kreĭn spaces. Up to now, all methods are based either on the matrix correction, or on non-convex minimization, or on feature-space embedding. Here we justify and evaluate a solution that uses the original (indefinite) similarity measure, in the original Kreĭn space. This solution is the result of a stabilization procedure. We establish the correspondence between the stabilization problem (which has to be solved) and a classical SVM based on minimization (which is easy to solve). We provide simple equations to go from one to the other (in both directions). This link between stabilization and minimization problems is the key to obtain a solution in the original Kreĭn space. Using KSVM, one can solve SVM with usually troublesome kernels (large negative eigenvalues or large numbers of negative eigenvalues). We show experiments showing that our algorithm KSVM outperforms all previously proposed approaches to deal with indefinite matrices in SVM-like kernel methods.

Many supervised learning problems involve learning from samples whose labels are corrupted in some way. For example, each sample may have some constant probability of being incorrectly labelled (learning with label noise), or one may have a pool of unlabelled samples in lieu of negative samples (learning from positive and unlabelled data). This paper uses class-probability estimation to study these and other corruption processes belonging to the mutually contaminated distributions framework (Scott et al., 2013), with three conclusions. First, one can optimise balanced error and AUC without knowledge of the corruption process parameters. Second, given estimates of the corruption parameters, one can minimise a range of classification risks. Third, one can estimate the corruption parameters using only corrupted data. Experiments confirm the efficacy of class-probability estimation in learning from corrupted labels.

A desired capability of deep learning is to understand the high-level, class-specific features via hierarchical features learning. However the training of deep architectures is costly comparing to simple shallow models. Bringing the high-level feature understanding into a simple shallow architecture remains an open question.

We propose a large margin method for asymmetric learning with ellipsoids, called eMIL, suited to multiple instance learning (MIL). We derive the distance between ellipsoids and the hyperplane, generalising the standard support vector machine. Negative bags in MIL contain only negative instances, and we treat them akin to uncertain observations in the robust optimisation framework. However, our method allows positive bags to cross the margin, since it is not known which instances within are positive. We show that representing bags as ellipsoids under the introduced distance is the most robust solution when treating a bag as a random variable with finite mean and covariance. Two algorithms are derived to solve the resulting non-convex optimization problem: a concave-convex procedure and a quasi-Newton method. Our method achieves competitive results on benchmark datasets. We introduce a MIL dataset from a real world application of detecting wheel defects from multiple partial observations, and show that eMIL outperforms competing approaches.

In many prediction problems, it is beneficial to obtain confidence estimates for the classification output. We consider the problem of estimating confidence sets in multiclass classification of real life datasets. Building on the theory of conformal predictors, we derive a class-conditional conformal predictor. This allows us to calibrate the confidence estimates in a class specific fashion, resulting in a more precise control of the prediction error rate for each class. We show that the class-conditional conformal predictor is asymptotically valid, and demonstrate that it indeed provides better calibration and efficiency on benchmark digit recognition datasets. In addition, we apply the class-conditional conformal predictor to a biological dataset for predicting localizations of proteins in order to demonstrate its performance in bioinformatics applications.

MOTIVATION: Biological systems are understood through iterations of modeling and experimentation. Not all experiments, however, are equally valuable for predictive modeling. This study introduces an efficient method for experimental design aimed at selecting dynamical models from data. Motivated by biological applications, the method enables the design of crucial experiments: it determines a highly informative selection of measurement readouts and time points. RESULTS: We demonstrate formal guarantees of design efficiency on the basis of previous results. By reducing our task to the setting of graphical models, we prove that the method finds a near-optimal design selection with a polynomial number of evaluations. Moreover, the method exhibits the best polynomial-complexity constant approximation factor, unless P = NP. We measure the performance of the method in comparison with established alternatives, such as ensemble non-centrality, on example models of different complexity. Efficient design accelerates the loop between modeling and experimentation: it enables the inference of complex mechanisms, such as those controlling central metabolic operation. AVAILABILITY: Toolbox 'NearOED' available with source code under GPL on the Machine Learning Open Source Software Web site (mloss.org).

How should we design experiments to maximize performance of a complex system, taking into account uncontrollable environmental conditions? How should we select relevant documents (ads) to display, given information about the user? These tasks can be formalized as contextual bandit problems, where at each round, we receive context (about the experimental conditions, the query), and have to choose an action (parameters, documents). The key challenge is to trade off exploration by gathering data for estimating the mean payoff function over the context-action space, and to exploit by choosing an action deemed optimal based on the gathered data. We model the payoff function as a sample from a Gaussian process defined over the joint context-action space, and develop CGP-UCB, an intuitive upper-confidence style algorithm. We show that by mixing and matching kernels for contexts and actions, CGP-UCB can handle a variety of practical applications. We further provide generic tools for deriving regret bounds when using such composite kernel functions. Lastly, we evaluate our algorithm on two case studies, in the context of automated vaccine design and sensor management. We show that context-sensitive optimization outperforms no or naive use of context.

Decoding models, such as those underlying multivariate classification algorithms, have been increasingly used to infer cognitive or clinical brain states from measures of brain activity obtained by functional magnetic resonance imaging (fMRI). The practicality of current classifiers, however, is restricted by two major challenges. First, due to the high data dimensionality and low sample size, algorithms struggle to separate informative from uninformative features, resulting in poor generalization performance. Second, popular discriminative methods such as support vector machines (SVMs) rarely afford mechanistic interpretability. In this paper, we address these issues by proposing a novel generative-embedding approach that incorporates neurobiologically interpretable generative models into discriminative classifiers. Our approach extends previous work on trial-by-trial classification for electrophysiological recordings to subject-by-subject classification for fMRI and offers two key advantages over conventional methods: it may provide more accurate predictions by exploiting discriminative information encoded in 'hidden' physiological quantities such as synaptic connection strengths; and it affords mechanistic interpretability of clinical classifications. Here, we introduce generative embedding for fMRI using a combination of dynamic causal models (DCMs) and SVMs. We propose a general procedure of DCM-based generative embedding for subject-wise classification, provide a concrete implementation, and suggest good-practice guidelines for unbiased application of generative embedding in the context of fMRI. We illustrate the utility of our approach by a clinical example in which we classify moderately aphasic patients and healthy controls using a DCM of thalamo-temporal regions during speech processing. Generative embedding achieves a near-perfect balanced classification accuracy of 98% and significantly outperforms conventional activation-based and correlation-based methods. This example demonstrates how disease states can be detected with very high accuracy and, at the same time, be interpreted mechanistically in terms of abnormalities in connectivity. We envisage that future applications of generative embedding may provide crucial advances in dissecting spectrum disorders into physiologically more well-defined subgroups.

Evaluating the performance of a classification algorithm critically requires a measure of the degree to which unseen examples have been identified with their correct class labels. In practice, generalizability is frequently estimated by averaging the accuracies obtained on individual cross-validation folds. This procedure, however, is problematic in two ways. First, it does not allow for the derivation of meaningful confidence intervals. Second, it leads to an optimistic estimate when a biased classifier is tested on an imbalanced dataset. We show that both problems can be overcome by replacing the conventional point estimate of accuracy by an estimate of the posterior distribution of the balanced accuracy.

The precision-recall curve (PRC) has become a widespread conceptual basis for assessing classification performance. The curve relates the positive predictive value of a classifier to its true positive rate and often provides a useful alternative to the well-known receiver operating characteristic (ROC). The empirical PRC, however, turns out to be a highly imprecise estimate of the true curve, especially in the case of a small sample size and class imbalance in favour of negative examples. Ironically, this situation tends to occur precisely in those applications where the curve would be most useful, e.g., in anomaly detection or information retrieval. Here, we propose to estimate the PRC on the basis of a simple distributional assumption about the decision values that generalizes the established binormal model for estimating smooth ROC curves. Using simulations, we show that our approach outperforms empirical estimates, and that an account of the class imbalance is crucial for obtaining unbiased PRC estimates.

This paper addresses the problem of active model selection for nonlinear dynamical systems. We propose a novel learning approach that selects the most informative subset of time-dependent variables for the purpose of Bayesian model inference. The model selection criterion maximizes the expected Kullback-Leibler divergence between the prior and the posterior probabilities over the models. The proposed strategy generalizes the standard D-optimal design, which is obtained from a uniform prior with Gaussian noise. In addition, our approach allows us to determine an information halting criterion for model identification. We illustrate the benefits of our approach by differentiating between 18 published biochemical models of the TOR signaling pathway, a model selection problem in systems biology. By generating pivotal selection experiments, our strategy outperforms the standard Aoptimal, D-optimal and E-optimal sequential design techniques.

Art. e1000173, 10 S.

In many applications it is desirable to learn from several kernels. "Multiple kernel learning" (MKL) allows the practitioner to optimize over linear combinations of kernels. By enforcing sparse coefficients, it also generalizes feature selection to kernel selection. We propose MKL for joint feature maps. This provides a convenient and principled way for MKL with multiclass problems. In addition, we can exploit the joint feature map to learn kernels on output spaces. We show the equivalence of several different primal formulations including different regularizers. We present several optimization methods, and compare a convex quadratically constrained quadratic program (QCQP) and two semi-infinite linear programs (SILPs) on toy data, showing that the SILPs are faster than the QCQP. We then demonstrate the utility of our method by applying the SILP to three real world datasets.

MOTIVATION: Despite many years of research on how to properly align sequences in the presence of sequencing errors, alternative splicing and micro-exons, the correct alignment of mRNA sequences to genomic DNA is still a challenging task. RESULTS: We present a novel approach based on large margin learning that combines accurate splice site predictions with common sequence alignment techniques. By solving a convex optimization problem, our algorithm-called PALMA-tunes the parameters of the model such that true alignments score higher than other alignments. We study the accuracy of alignments of mRNAs containing artificially generated micro-exons to genomic DNA. In a carefully designed experiment, we show that our algorithm accurately identifies the intron boundaries as well as boundaries of the optimal local alignment. It outperforms all other methods: for 5702 artificially shortened EST sequences from Caenorhabditis elegans and human, it correctly identifies the intron boundaries in all except two cases. The best other method is a recently proposed method called exalin which misaligns 37 of the sequences. Our method also demonstrates robustness to mutations, insertions and deletions, retaining accuracy even at high noise levels. AVAILABILITY: Datasets for training, evaluation and testing, additional results and a stand-alone alignment tool implemented in C++ and python are available at http://www.fml.mpg.de/raetsch/projects/palma

This paper addresses the problem of choosing a kernel suitable for estimation with a support vector machine, hence further automating machine learning. This goal is achieved by defining a reproducing kernel Hilbert space on the space of kernels itself. Such a formulation leads to a statistical estimation problem similar to the problem of minimizing a regularized risk functional. We state the equivalent representer theorem for the choice of kernels and present a semidefinite programming formulation of the resulting optimization problem. Several recipes for constructing hyperkernels are provided, as well as the details of common machine learning problems. Experimental results for classification, regression and novelty detection on UCI data show the feasibility of our approach.

In this paper we show that many kernel methods can be adapted to deal with indefinite kernels, that is, kernels which are not positive semidefinite. They do not satisfy Mercer's condition and they induce associated functional spaces called Reproducing Kernel Kreĭn Spaces (RKKS), a generalization of Reproducing Kernel Hilbert Spaces (RKHS).Machine learning in RKKS shares many "nice" properties of learning in RKHS, such as orthogonality and projection. However, since the kernels are indefinite, we can no longer minimize the loss, instead we stabilize it. We show a general representer theorem for constrained stabilization and prove generalization bounds by computing the Rademacher averages of the kernel class. We list several examples of indefinite kernels and investigate regularization methods to solve spline interpolation. Some preliminary experiments with indefinite kernels for spline smoothing are reported for truncated spectral factorization, Landweber-Fridman iterations, and MR-II.

We consider the problem of choosing a kernel suitable for estimation using a Gaussian Process estimator or a Support Vector Machine. A novel solution is presented which involves defining a Reproducing Ker- nel Hilbert Space on the space of kernels itself. By utilizing an analog of the classical representer theorem, the problem of choosing a kernel from a parameterized family of kernels (e.g. of varying width) is reduced to a statistical estimation problem akin to the problem of minimizing a regularized risk functional. Various classical settings for model or kernel selection are special cases of our framework.

Password authentication is the most commonly used identification system in today's computer world. Its security can be enhanced using typing biometrics as a transparent layer of user authentication. Our research focuses on using the time period between keystrokes as the measure of the individual's typing pattern. The typing pattern of a particular individual can be represented by the weights of a fully trained multilayer perceptron. Alternatively, each user's typing pattern can be viewed as a cluster of measurements that can be differentiated from clusters of other users.