Developers now have access to a growing array of increasingly autonomous AI tools for software development. While many studies examine copilots that provide chat assistance or code completions, evaluations of coding agents—which can automatically write files and run code—still rely on static benchmarks. We present the first controlled study of developer interactions with coding agents, characterizing how more autonomous AI tools affect productivity and experience. We evaluate two leading copilot and agentic coding assistants, recruiting participants who regularly use the former. Our results show agents can assist developers in ways that surpass copilots (e.g., completing tasks humans may not have accomplished) and reduce the effort required to finish tasks. Yet challenges remain for broader adoption, including ensuring users adequately understand agent behaviors. Our findings reveal how workflows shift with coding agents and how interactions differ from copilots, motivating recommendations for researchers and highlighting challenges in adopting agentic systems.

Individual human decision-makers may benefit from different forms of support to improve decision outcomes, but when will each form of support yield better outcomes? In this work, we posit that personalizing access to decision support tools can be an effective mechanism for instantiating the appropriate use of AI assistance. Specifically, we propose the general problem of learning a decision support policy that, for a given input, chooses which form of support to provide to decision-makers for whom we initially have no prior information. We develop Modiste, an interactive tool to learn personalized decision support policies. Modiste leverages stochastic contextual bandit techniques to personalize a decision support policy for each decision-maker. In our computational experiments, we characterize the expertise profiles of decision-makers for whom personalized policies will outperform offline policies, including population-wide baselines. Our experiments include realistic forms of support (e.g., expert consensus and predictions from a large language model) on vision and language tasks. Our human subject experiments add nuance to and bolster our computational experiments, demonstrating the practical utility of personalized policies when real users benefit from accessing support across tasks.

Optimization in deep learning remains poorly understood. A key difficulty is that optimizers exhibit complex oscillatory dynamics, referred to as "edge of stability," which cannot be captured by traditional optimization theory. In this paper, we show that the path taken by an oscillatory optimizer can often be captured by a _central flow_: a differential equation which directly models the time-averaged (i.e. smoothed) optimization trajectory. We empirically show that these central flows can predict long-term optimization trajectories for generic neural networks with a high degree of numerical accuracy. By interpreting these flows, we are able to understand how gradient descent makes progress even as the loss sometimes goes up; how adaptive optimizers ``adapt'' to the local loss landscape; and how adaptive optimizers implicitly seek out regions of weight space where they can take larger steps. These insights (and others) are not apparent from the optimizers' update rules, but are revealed by the central flows. Therefore, we believe that central flows constitute a promising tool for reasoning about optimization in deep learning.

Following its success for vision and text, the "foundation model" (FM) paradigm—pretraining large models on massive data, then fine-tuning on target tasks—has rapidly expanded to domains in the sciences, engineering, healthcare, and beyond. Has this achieved what the original FMs accomplished, i.e. the supplanting of traditional supervised learning in their domains? To answer we look at three modalities—genomics, satellite imaging, and time series—with multiple recent FMs and compare them to a standard supervised learning workflow: model development, hyperparameter tuning, and training, all using only data from the target task. Across these three specialized domains, we find that it is consistently possible to train simple supervised models—no more complicated than a lightly modified wide ResNet or UNet—that match or even outperform the latest foundation models. Our work demonstrates that the benefits of large-scale pretraining have yet to be realized in many specialized areas, reinforces the need to compare new FMs to strong, well-tuned baselines, and introduces two new, easy-to-use, open-source, and automated workflows for doing so.

Evaluating in-the-wild coding capabilities of large language models (LLMs) is a challenging endeavor with no existing solution. We introduce Copilot Arena, a platform to collect user preferences through native integration into a developer's working environment. Copilot Arena comprises a novel interface for comparing pairs of model outputs, a sampling strategy to reduce experienced latency, and a prompting scheme to enable code completion functionality. Copilot Arena has served over 4.5 million suggestions from 10 models and collected over 11k pairwise judgements. Our results highlight the importance of model evaluations in integrated settings. We find that model rankings from Copilot Arena differ from those of existing evaluations, which we attribute to the unique distribution of data and tasks contained in Copilot Arena. We also identify novel insights into human preferences on code such as an observed consistency in user preference across programming languages yet significant variation in preference due to task category. We open-source Copilot Arena and release data to enable human-centric evaluations and improve understanding of coding assistants.

We introduce Toto, a time series forecasting foundation model with 151 million parameters. Toto uses a modern decoder-only architecture coupled with architectural innovations designed to account for specific challenges found in multivariate observability time series data. Toto's pre-training corpus is a mixture of observability data, open datasets, and synthetic data, and is 4-10$\times$ larger than those of leading time series foundation models. Additionally, we introduce BOOM, a large-scale benchmark consisting of 350 million observations across 2,807 real-world time series. For both Toto and BOOM, we source observability data exclusively from our own telemetry and internal observability metrics. Extensive evaluations demonstrate that Toto achieves state-of-the-art performance on both BOOM and on established general purpose time series forecasting benchmarks. Toto's model weights, inference code, and evaluation scripts, as well as BOOM's data and evaluation code, are all available as open source under the Apache 2.0 License.

Test-time scaling in agentic tasks often relies on generating long reasoning traces ("think" more) before acting, but this does not allow agents to acquire new information from the environment or adapt behavior over time. In this work, we propose scaling test-time interaction, an untapped dimension for test-time scaling that increases the agent's interaction horizon to enable rich behaviors such as exploration, backtracking, and dynamic re-planning within a single rollout. To demonstrate the promise of this scaling dimension, we situate our study in the domain of web agents. We first show that even prompting-based interaction scaling can improve task success on web benchmarks non-trivially. Building on this, we introduce TTI, a curriculum-based online reinforcement learning (RL) approach that trains agents by adaptively adjusting their interaction lengths during rollout. Using a Gemma 3 12B model, TTI sets a new state-of-the-art among open-source agents trained on public data on WebVoyager and WebArena. Case studies further reveal that TTI enables agents to balance exploration and exploitation adaptively. Our results establish interaction scaling as a powerful, complementary axis to scaling per-action compute, offering new avenues for training robust and adaptive agents.

Most existing evaluations of explainable machine learning (ML) methods rely on simplifying assumptions or proxies that do not reflect real-world use cases; the handful of more robust evaluations on real-world settings have shortcomings in their design, generally leading to overestimation of methods' real-world utility. In this work, we seek to address this by conducting a study that evaluates post-hoc explainable ML methods in a setting consistent with the application context and provide a template for future evaluation studies. We modify and improve a prior study on e-commerce fraud detection by relaxing the original work's simplifying assumptions that departed from the deployment context. Our study finds no evidence for the utility of the tested explainable ML methods in the context, which is a drastically different conclusion from the earlier work. This highlights how seemingly trivial experimental design choices can yield misleading conclusions about method utility. In addition, our work carries lessons about the necessity of not only evaluating explainable ML methods using tasks, data, users, and metrics grounded in the intended application context but also developing methods tailored to specific applications, moving beyond general-purpose explainable ML methods.

Solving a linear system ${\bf Ax}={\bf b}$ is a fundamental scientific computing primitive for which numerous solvers and preconditioners have been developed. These come with parameters whose optimal values depend on the system being solved and are often impossible or too expensive to identify; thus in practice sub-optimal heuristics are used. We consider the common setting in which many related linear systems need to be solved, e.g. during a single numerical simulation. In this scenario, can we sequentially choose parameters that attain a near-optimal overall number of iterations, without extra matrix computations? We answer in the affirmative for Successive Over-Relaxation (SOR), a standard solver whose parameter $\omega$ has a strong impact on its runtime. For this method, we prove that a bandit online learning algorithm—using only the number of iterations as feedback—can select parameters for a sequence of instances such that the overall cost approaches that of the best fixed $\omega$ as the sequence length increases. Furthermore, when given additional structural information, we show that a _contextual_ bandit method asymptotically achieves the performance of the _instance-optimal_ policy, which selects the best $\omega$ for each instance. Our work provides the first learning-theoretic treatment of high-precision linear system solvers and the first end-to-end guarantees for data-driven scientific computing, demonstrating theoretically the potential to speed up numerical methods using well-understood learning algorithms.

Machine learning models with high accuracy on test data can still produce systematic failures, such as harmful biases and safety issues, when deployed in the real world. To detect and mitigate such failures, practitioners run behavioral evaluation of their models, checking model outputs for specific types of inputs. Behavioral evaluation is important but challenging, requiring that practitioners discover real-world patterns and validate systematic failures. We conducted 18 semi-structured interviews with ML practitioners to better understand the challenges of behavioral evaluation and found that it is a collaborative, use-case-first process that is not adequately supported by existing task- and domain-specific tools. Using these findings, we designed zeno, a general-purpose framework for visualizing and testing AI systems across diverse use cases. In four case studies with participants using zeno on real-world models, we found that practitioners were able to reproduce previous manual analyses and discover new systematic failures.

Fine-tuning large-scale pretrained models has led to tremendous progress in well-studied modalities such as vision and NLP. However, similar gains have not been observed in many other modalities due to a lack of relevant pretrained models. In this work, we propose ORCA, a general cross-modal fine-tuning framework that extends the applicability of a single large-scale pretrained model to diverse modalities. ORCA adapts to a target task via an align-then-refine workflow: given the target input, ORCA first learns an embedding network that aligns the embedded feature distribution with the pretraining modality. The pretrained model is then fine-tuned on the embedded data to exploit the knowledge shared across modalities. Through extensive experiments, we show that ORCA obtains state-of-the-art results on 3 benchmarks containing over 60 datasets from 12 modalities, outperforming a wide range of hand-designed, AutoML, general-purpose, and task-specific cross-modal methods. We highlight the importance of data alignment via a series of ablation studies and exemplify ORCA's utility in data-limited regimes.

Saliency methods are a popular class of feature attribution explanation methods that aim to capture a model’s predictive reasoning by identifying "important" pixels in an input image. However, the development and adoption of these methods are hindered by the lack of access to ground-truth model reasoning, which prevents accurate evaluation. In this work, we design a synthetic benchmarking framework, SMERF, that allows us to perform ground-truth-based evaluation while controlling the complexity of the model’s reasoning. Experimentally, SMERF reveals significant limitations in existing saliency methods and, as a result, represents a useful tool for the development of new saliency methods.

An important unresolved challenge in the theory of regularization is to set the regularization coefficients of popular techniques like the ElasticNet with general provable guarantees. We consider the problem of tuning the regularization parameters of Ridge regression, LASSO, and the ElasticNet across multiple problem instances, a setting that encompasses both cross-validation and multi-task hyperparameter optimization. We obtain a novel structural result for the ElasticNet which characterizes the loss as a function of the tuning parameters as a piecewise-rational function with algebraic boundaries. We use this to bound the structural complexity of the regularized loss functions and show generalization guarantees for tuning the ElasticNet regression coefficients in the statistical setting. We also consider the more challenging online learning setting, where we show vanishing average expected regret relative to the optimal parameter pair. We further extend our results to tuning classification algorithms obtained by thresholding regression fits regularized by Ridge, LASSO, or ElasticNet. Our results are the first general learning-theoretic guarantees for this important class of problems that avoid strong assumptions on the data distribution. Furthermore, our guarantees hold for both validation and popular information criterion objectives.

A growing body of research runs human subject evaluations to study whether providing users with explanations of machine learning models can help them with practical real-world use cases. However, running user studies is challenging and costly, and consequently each study typically only evaluates a limited number of different settings, e.g., studies often only evaluate a few arbitrarily selected model explanation methods. To address these challenges and aid user study design, we introduce Simulated Evaluations (SimEvals). SimEvals involve training algorithmic agents that take as input the information content (such as model explanations) that would be presented to the user, to predict answers to the use case of interest. The algorithmic agent's test set accuracy provides a measure of the predictiveness of the information content for the downstream use case. We run a comprehensive evaluation on three real-world use cases (forward simulation, model debugging, and counterfactual reasoning) to demonstrate that SimEvals can effectively identify which explanation methods will help humans for each use case. These results provide evidence that \simevals{} can be used to efficiently screen an important set of user study design decisions, e.g., selecting which explanations should be presented to the user, before running a potentially costly user study.

When subjected to automated decision-making, decision subjects may strategically modify their observable features in ways they believe will maximize their chances of receiving a favorable decision. In many practical situations, the underlying assessment rule is deliberately kept secret to avoid gaming and maintain competitive advantage. The resulting opacity forces the decision subjects to rely on incomplete information when making strategic feature modifications. We capture such settings as a game of Bayesian persuasion, in which the decision maker offers a form of recourse to the decision subject by providing them with an action recommendation (or signal) to incentivize them to modify their features in desirable ways. We show that when using persuasion, the decision maker and decision subject are never worse off in expectation, while the decision maker can be significantly better off. While the decision maker’s problem of finding the optimal Bayesian incentive compatible (BIC) signaling policy takes the form of optimization over infinitely many variables, we show that this optimization can be cast as a linear program over finitely-many regions of the space of possible assessment rules. While this reformulation simplifies the problem dramatically, solving the linear program requires reasoning about exponentially-many variables, even in relatively simple cases. Motivated by this observation, we provide a polynomial-time approximation scheme that recovers a near-optimal signaling policy. Finally, our numerical simulations on semi-synthetic data empirically demonstrate the benefits of using persuasion in the algorithmic recourse setting.

A burgeoning paradigm in algorithm design is the field of algorithms with predictions, in which algorithms can take advantage of a possibly-imperfect prediction of some aspect of the problem. While much work has focused on using predictions to improve competitive ratios, running times, or other performance measures, less effort has been devoted to the question of how to obtain the predictions themselves, especially in the critical online setting. We introduce a general design approach for algorithms that learn predictors: (1) identify a functional dependence of the performance measure on the prediction quality and (2) apply techniques from online learning to learn predictors, tune robustness-consistency trade-offs, and bound the sample complexity. We demonstrate the effectiveness of our approach by applying it to bipartite matching, ski-rental, page migration, and job scheduling. In several settings we improve upon multiple existing results while utilizing a much simpler analysis, while in the others we provide the first learning-theoretic guarantees.

While neural architecture search (NAS) has enabled automated machine learning (AutoML) for well-researched areas, its application to tasks beyond computer vision is still under-explored. As less-studied domains are precisely those where we expect AutoML to have the greatest impact, in this work we study NAS for efficiently solving diverse problems. Seeking an approach that is fast, simple, and broadly applicable, we fix a standard convolutional network (CNN) topology and propose to search for the right kernel sizes and dilations its operations should take on. This dramatically expands the model's capacity to extract features at multiple resolutions for different types of data while only requiring search over the operation space. To overcome the efficiency challenges of naive weight-sharing in this search space, we introduce DASH, a differentiable NAS algorithm that computes the mixture-of-operations using the Fourier diagonalization of convolution, achieving both a better asymptotic complexity and an up-to-10x search time speedup in practice. We evaluate DASH on ten tasks spanning a variety of application domains such as PDE solving, protein folding, and heart disease detection. DASH outperforms state-of-the-art AutoML methods in aggregate, attaining the best-known automated performance on seven tasks. Meanwhile, on six of the ten tasks, the combined search and retraining time is less than 2x slower than simply training a CNN backbone that is far less accurate.

Most existing neural architecture search (NAS) benchmarks and algorithms prioritize well-studied tasks, e.g. image classification on CIFAR or ImageNet. This makes the performance of NAS approaches in more diverse areas poorly understood. In this paper, we present NAS-Bench-360, a benchmark suite to evaluate methods on domains beyond those traditionally studied in architecture search, and use it to address the following question: do state-of-the-art NAS methods perform well on diverse tasks? To construct the benchmark, we curate ten tasks spanning a diverse array of application domains, dataset sizes, problem dimensionalities, and learning objectives. Each task is carefully chosen to interoperate with modern CNN-based search methods while possibly being far-afield from its original development domain. To speed up and reduce the cost of NAS research, for two of the tasks we release the precomputed performance of 15,625 architectures comprising a standard CNN search space. Experimentally, we show the need for more robust NAS evaluation of the kind NAS-Bench-360 enables by showing that several modern NAS procedures perform inconsistently across the ten tasks, with many catastrophically poor results. We also demonstrate how NAS-Bench-360 and its associated precomputed results will enable future scientific discoveries by testing whether several recent hypotheses promoted in the NAS literature hold on diverse tasks. NAS-Bench-360 is hosted at https://nb360.ml.cmu.edu.

Meta-learning has enabled learning statistical models that can be quickly adapted to new prediction tasks. Motivated by use-cases in personalized federated learning, we study the often overlooked aspect of the modern meta-learning algorithms—their data efficiency. To shed more light on which methods are more efficient, we use techniques from algorithmic stability to derive bounds on the transfer risk that have important practical implications, indicating how much supervision is needed and how it must be allocated for each method to attain the desired level of generalization. Further, we introduce a new simple framework for evaluating meta-learning methods under a limit on the available supervision, conduct an empirical study of MAML, Reptile, andProtoNets, and demonstrate the differences in the behavior of these methods on few-shot and federated learning benchmarks. Finally, we propose active meta-learning, which incorporates active data selection into learning-to-learn, leading to better performance of all methods in the limited supervision regime.

We analyze the meta-learning of the initialization and step-size of learning algorithms for piecewise-Lipschitz functions, a non-convex setting with applications to both machine learning and algorithms. Starting from recent regret bounds for the exponential forecaster on losses with dispersed discontinuities, we generalize them to be initialization-dependent and then use this result to propose a practical meta-learning procedure that learns both the initialization and the step-size of the algorithm from multiple online learning tasks. Asymptotically, we guarantee that the average regret across tasks scales with a natural notion of task-similarity that measures the amount of overlap between near-optimal regions of different tasks. Finally, we instantiate the method and its guarantee in two important settings: robust meta-learning and multi-task data-driven algorithm design.

Tuning hyperparameters is a crucial but arduous part of the machine learning pipeline. Hyperparameter optimization is even more challenging in federated learning, where models are learned over a distributed network of heterogeneous devices; here, the need to keep data on device and perform local training makes it difficult to efficiently train and evaluate configurations. In this work, we investigate the problem of federated hyperparameter tuning. We first identify key challenges and show how standard approaches may be adapted to form baselines for the federated setting. Then, by making a novel connection to the neural architecture search technique of weight-sharing, we introduce a new method, FedEx, to accelerate federated hyperparameter tuning that is applicable to widely-used federated optimization methods such as FedAvg and recent variants. Theoretically, we show that a FedEx variant correctly tunes the on-device learning rate in the setting of online convex optimization across devices. Empirically, we show that FedEx can outperform natural baselines for federated hyperparameter tuning by several percentage points on the Shakespeare, FEMNIST, and CIFAR-10 benchmarks—obtaining higher accuracy using the same training budget.

An important goal of AutoML is to automate-away the design of neural networks on new tasks in under-explored domains. Motivated by this goal, we study the problem of enabling users to discover the right neural operations given data from their specific domain. We introduce a search space of operations called XD-Operations that mimic the inductive bias of standard multi-channel convolutions while being much more expressive: we prove that it includes many named operations across multiple application areas. Starting with any standard backbone such as ResNet, we show how to transform it into a search space over XD-operations and how to traverse the space using a simple weight sharing scheme. On a diverse set of tasks—solving PDEs, distance prediction for protein folding, and music modeling—our approach consistently yields models with lower error than baseline networks and often even lower error than expert-designed domain-specific approaches.

Fair representations are a powerful tool for establishing criteria like statistical parity, proxy non-discrimination, and equality of opportunity in learned models. Existing techniques for learning these representations are typically model-agnostic, as they preprocess the original data such that the output satisfies some fairness criterion, and can be used with arbitrary learning methods. In contrast, we demonstrate the promise of learning a model-aware fair representation, focusing on kernel-based models. We leverage the classical Sufficient Dimension Reduction (SDR) framework to construct representations as subspaces of the reproducing kernel Hilbert space (RKHS), whose member functions are guaranteed to satisfy fairness. Our method supports several fairness criteria, continuous and discrete data, and multiple protected attributes. We further show how to calibrate the accuracy tradeoff by characterizing it in terms of the principal angles between subspaces of the RKHS. Finally, we apply our approach to obtain the first Fair Gaussian Process (FGP) prior for fair Bayesian learning, and show that it is competitive with, and in some cases outperforms, state-of-the-art methods on real data.

Group fairness, a class of fairness notions that measure how different groups of individuals are treated differently according to their protected attributes, has been shown to conflict with one another, often with a necessary cost in loss of model’s predictive performance. We propose a general diagnostic that enables systematic characterization of these trade-offs in group fairness. We observe that the majority of group fairness notions can be expressed via the fairness-confusion tensor, which is the confusion matrix split according to the protected attribute values. We frame several optimization problems that directly optimize both accuracy and fairness objectives over the elements of this tensor, which yield a general perspective for understanding multiple trade-offs including group fairness incompatibilities. It also suggests an alternate post-processing method for designing fair classifiers. On synthetic and real datasets, we demonstrate the use cases of our diagnostic, particularly on understanding the trade-off landscape between accuracy and fairness.

A common workflow in data exploration is to learn a low-dimensional representation of the data, identify groups of points in that representation, and examine the differences between the groups to determine what they represent. We treat this workflow as an interpretable machine learning problem by leveraging the model that learned the low-dimensional representation to help identify the key differences between the groups. To solve this problem, we introduce a new type of explanation, a Global Counterfactual Explanation (GCE), and our algorithm, Transitive Global Translations (TGT), for computing GCEs. TGT identifies the differences between each pair of groups using compressed sensing but constrains those pairwise differences to be consistent among all of the groups. Empirically, we demonstrate that TGT is able to identify explanations that accurately explain the model while being relatively sparse, and that these explanations match real patterns in the data.

Most of the work on interpretable machine learning has focused on designing either inherently interpretable models, which typically trade-off accuracy for interpretability, or post-hoc explanation systems, whose explanation quality can be unpredictable. Our method, ExpO, is a hybridization of these approaches that regularizes a model for explanation quality at training time. Importantly, these regularizers are differentiable, model agnostic, and require no domain knowledge to define. We demonstrate that post-hoc explanations for ExpO-regularized models have better explanation quality, as measured by the common fidelity and stability metrics. We verify that improving these metrics leads to significantly more useful explanations with a user study on a realistic task.

We study the problem of meta-learning through the lens of online convex optimization, developing a meta-algorithm bridging the gap between popular gradient-based meta-learning and classical regularization-based multi-task transfer methods. Our method is the first to simultaneously satisfy good sample efficiency guarantees in the convex setting, with generalization bounds that improve with task-similarity, while also being computationally scalable to modern deep learning architectures and the many-task setting. Despite its simplicity, the algorithm matches, up to a constant factor, a lower bound on the performance of any such parameter-transfer method under natural task similarity assumptions. We use experiments in both convex and deep learning settings to verify and demonstrate the applicability of our theory.

We build a theoretical framework for designing and understanding practical meta-learning methods that integrates sophisticated formalizations of task-similarity with the extensive literature on online convex optimization and sequential prediction algorithms. Our approach enables the task-similarity to be learned adaptively, provides sharper transfer-risk bounds in the setting of statistical learning-to-learn, and leads to straightforward derivations of average-case regret bounds for efficient algorithms in settings where the task-environment changes dynamically or the tasks share a certain geometric structure. We use our theory to modify several popular meta-learning algorithms and improve their training and meta-test-time performance on standard problems in few-shot and federated learning.

Model interpretability is an increasingly important component of practical machine learning. Some of the most common forms of interpretability systems are example-based, local, and global explanations. One of the main challenges in interpretability is designing explanation systems that can capture aspects of each of these explanation types, in order to develop a more thorough understanding of the model. We address this challenge in a novel model called MAPLE that uses local linear modeling techniques along with a dual interpretation of random forests (both as a supervised neighborhood approach and as a feature selection method). MAPLE has two fundamental advantages over existing interpretability systems. First, while it is effective as a black-box explanation system, MAPLE itself is a highly accurate predictive model that provides faithful self explanations, and thus sidesteps the typical accuracy-interpretability trade-off. Specifically, we demonstrate, on several UCI datasets, that MAPLE is at least as accurate as random forests and that it produces more faithful local explanations than LIME, a popular interpretability system. Second, MAPLE provides both example-based and local explanations and can detect global patterns, which allows it to diagnose limitations in its local explanations.

Industrial-scale machine learning applications often train and maintain massive models that can be on the order of hundreds of millions to billions of parameters. Model parallelism thus plays a significant role to support these machine learning tasks. Recent work in this area has been dominated by parameter server architectures that follow an asynchronous computation model, introducing added complexity and approximation in order to scale to massive workloads. In this work, we explore model parallelism in the distributed bulk-synchronous parallel (BSP) setting, leveraging some recent progress made in the area of high performance computing, in order to address these complexity and approximation issues. Using collaborative filtering as a case-study, we introduce an efficient model parallel industrial scale algorithm for alternating least squares (ALS), along with a highly optimized implementation of ALS that serves as the default implementation in MLlib, Apache Spark's machine learning library. Our extensive empirical evaluation demonstrates that our implementation in MLlib compares favorably to the leading open-source parameter server framework, and our implementation scales to massive problems on the order of 50 billion ratings and close to 1 billion parameters.

Decision trees and random forests are well established models that not only offer good predictive performance, but also provide rich feature importance information. While practitioners often employ variable importance methods that rely on this impurity-based information, these methods remain poorly characterized from a theoretical perspective. We provide novel insights into the performance of these methods by deriving finite sample performance guarantees in a high-dimensional setting under various modeling assumptions. We further demonstrate the effectiveness of these impurity-based methods via an extensive set of simulations.

Federated learning poses new statistical and systems challenges in training machine learning models over distributed networks of devices. In this work, we show that multi-task learning is naturally suited to handle the statistical challenges of this setting, and propose a novel systems-aware optimization method, MOCHA, that is robust to practical systems issues. Our method and theory for the first time consider issues of high communication cost, stragglers, and fault tolerance for distributed multi-task learning. The resulting method achieves significant speedups compared to alternatives in the federated setting, as we demonstrate through simulations on real-world federated datasets.

Techniques for nonparametric regression based on fitting small-scale local models at prediction time have long been studied in statistics and pattern recognition, but have received less attention in modern large-scale machine learning applications. In practice, such methods are generally applied to low-dimensional problems, but may falter with high-dimensional predictors if they use a Euclidean distance-based kernel. We propose a new method, SILO, for fitting prediction-time local models that uses supervised neighborhoods that adapt to the local shape of the regression surface. To learn such neighborhoods, we use a weight function between points derived from random forests. We prove the consistency of SILO, and demonstrate through simulations and real data that our method works well in both the serial and distributed settings. In the latter case, SILO learns the weighting function in a divide-and-conquer manner, entirely avoiding communication at training time.

Motivated by the task of hyperparameter optimization, we introduce the \em non-stochastic best-arm identification problem. We identify an attractive algorithm for this setting that makes no assumptions on the convergence behavior of the arms’ losses, has no free-parameters to adjust, provably outperforms the uniform allocation baseline in favorable conditions, and performs comparably (up to \log factors) otherwise. Next, by leveraging the iterative nature of many learning algorithms, we cast hyperparameter optimization as an instance of non-stochastic best-arm identification. Our empirical results show that, by allocating more resources to promising hyperparameter settings, our approach achieves comparable test accuracies an order of magnitude faster than the uniform strategy. The robustness and simplicity of our approach makes it well-suited to ultimately replace the uniform strategy currently used in most machine learning software packages.

Deep distributed decision trees and tree ensembles have grown in importance due to the need to model increasingly large datasets. However, PLANET, the standard distributed tree learning algorithm implemented in systems such as \xgboost and Spark MLlib, scales poorly as data dimensionality and tree depths grow. We present Yggdrasil, a new distributed tree learning method that outperforms existing methods by up to 24x. Unlike PLANET, Yggdrasil is based on vertical partitioning of the data (i.e., partitioning by feature), along with a set of optimized data structures to reduce the CPU and communication costs of training. Yggdrasil (1) trains directly on compressed data for compressible features and labels; (2) introduces efficient data structures for training on uncompressed data; and (3) minimizes communication between nodes by using sparse bitvectors. Moreover, while PLANET approximates split points through feature binning, Yggdrasil does not require binning, and we analytically characterize the impact of this approximation. We evaluate Yggdrasil against the MNIST 8M dataset and a high-dimensional dataset at Yahoo; for both, Yggdrasil is faster by up to an order of magnitude.

Apache Spark is a popular open-source platform for large-scale data processing that is well-suited for iterative machine learning tasks. In this paper we present MLlib, Spark's open- source distributed machine learning library. MLlib provides efficient functionality for a wide range of learning settings and includes several underlying statistical, optimization, and linear algebra primitives. Shipped with Spark, MLlib supports several languages and provides a high-level API that leverages Spark's rich ecosystem to simplify the development of end-to-end machine learning pipelines. MLlib has experienced a rapid growth due to its vibrant open-source community of over 140 contributors, and includes extensive documentation to support further growth and to let users quickly get up to speed.

The proliferation of massive datasets combined with the development of sophisticated analytical techniques has enabled a wide variety of novel applications such as improved product recommendations, automatic image tagging, and improved speech-driven interfaces. A major obstacle to supporting these predictive applications is the challenging and expensive process of identifying and training an appropriate predictive model. Recent efforts aiming to automate this process have focused on single node implementations and have assumed that model training itself is a black box, limiting their usefulness for applications driven by large-scale datasets. In this work, we build upon these recent efforts and propose an architecture for automatic machine learning at scale comprised of a cost-based cluster resource allocation estimator, advanced hyper-parameter tuning techniques, bandit resource allocation via runtime algorithm introspection, and physical optimization via batching and optimal resource allocation. The result is TuPAQ, a component of the MLbase system that automatically finds and trains models for a user's predictive application with comparable quality to those found using exhaustive strategies, but an order of magnitude more efficiently than the standard baseline approach. TuPAQ scales to models trained on Terabytes of data across hundreds of machines.

Modern data analytics applications typically process massive amounts of data on clusters of tens, hundreds, or thousands of machines to support near-real-time decisions.The quantity of data and limitations of disk and memory bandwidth often make it infeasible to deliver answers at interactive speeds. However, it has been widely observed that many applications can tolerate some degree of inaccuracy. This is especially true for exploratory queries on data, where users are satisfied with "close-enough" answers if they can come quickly. A popular technique for speeding up queries at the cost of accuracy is to execute each query on a sample of data, rather than the whole dataset. To ensure that the returned result is not too inaccurate, past work on approximate query processing has used statistical techniques to estimate "error bars" on returned results. However, existing work in the sampling-based approximate query processing (S-AQP) community has not validated whether these techniques actually generate accurate error bars for real query workloads. In fact, we find that error bar estimation often fails on real world production workloads. Fortunately, it is possible to quickly and accurately diagnose the failure of error estimation for a query. In this paper, we show that it is possible to implement a query approximation pipeline that produces approximate answers and reliable error bars at interactive speeds.

MOTIVATION: Computational methods are essential to extract actionable information from raw sequencing data, and to thus fulfill the promise of next-generation sequencing technology. Unfortunately, computational tools developed to call variants from human sequencing data disagree on many of their predictions, and current methods to evaluate accuracy and computational performance are ad hoc and incomplete. Agreement on benchmarking variant calling methods would stimulate development of genomic processing tools and facilitate communication among researchers. RESULTS: We propose SMaSH, a benchmarking methodology for evaluating germline variant calling algorithms. We generate synthetic datasets, organize and interpret a wide range of existing benchmarking data for real genomes and propose a set of accuracy and computational performance metrics for evaluating variant calling methods on these benchmarking data. Moreover, we illustrate the utility of SMaSH to evaluate the performance of some leading single-nucleotide polymorphism, indel and structural variant calling algorithms. AVAILABILITY AND IMPLEMENTATION: We provide free and open access online to the SMaSH tool kit, along with detailed documentation, at smash.cs.berkeley.edu

The effects of social influence and homophily suggest that both network structure and node-attribute information should inform the tasks of link prediction and node-attribute inference. Recently, Yin et al. [2010a, 2010b] proposed an attribute-augmented social network model, which we callSocial-Attribute Network(SAN), to integrate network structure and node attributes to perform both link prediction and attribute inference. They focused on generalizing the random walk with a restart algorithm to the SAN framework and showed improved performance. In this article, we extend the SAN framework with several leading supervised and unsupervised link-prediction algorithms and demonstrate performance improvement for each algorithm on both link prediction and attribute inference. Moreover, we make the novel observation that attribute inference can help inform link prediction, that is, link-prediction accuracy is further improved by first inferring missing attributes. We comprehensively evaluate these algorithms and compare them with other existing algorithms using a novel, large-scale Google+ dataset, which we make publicly available (http://www.cs.berkeley.edu/~stevgong/gplus.html).

Vision problems ranging from image clustering to motion segmentation to semi-supervised learning can naturally be framed as subspace segmentation problems, in which one aims to recover multiple low-dimensional subspaces from noisy and corrupted input data. Low-Rank Representation (LRR), a convex formulation of the subspace segmentation problem, is provably and empirically accurate on small problems but does not scale to the massive sizes of modern vision datasets. Moreover, past work aimed at scaling up low-rank matrix factorization is not applicable to LRR given its non-decomposable constraints. In this work, we propose a novel divide-and-conquer algorithm for large-scale subspace segmentation that can cope with LRR's non-decomposable constraints and maintains LRR's strong recovery guarantees. This has immediate implications for the scalability of subspace segmentation, which we demonstrate on a benchmark face recognition dataset and in simulations. We then introduce novel applications of LRR-based subspace segmentation to large-scale semi-supervised learning for multimedia event detection, concept detection, and image tagging. In each case, we obtain state-of-the-art results and order-of-magnitude speed ups.

MLI is an Application Programming Interface designed to address the challenges of building Machine Learning algorithms in a distributed setting based on data-centric computing. Its primary goal is to simplify the development of high-performance, scalable, distributed algorithms. Our initial results show that, relative to existing systems, this interface can be used to build distributed implementations of a wide variety of common Machine Learning algorithms with minimal complexity and highly competitive performance and scalability.

As datasets become larger, more complex, and more available to diverse groups of analysts, it would be quite useful to be able to automatically and generically assess the quality of estimates, much as we are able to automatically train and evaluate predictive models such as classifiers. However, despite the fundamental importance of estimator quality assessment in data analysis, this task has eluded highly automatic solutions. While the bootstrap provides perhaps the most promising step in this direction, its level of automation is limited by the difficulty of evaluating its finite sample performance and even its asymptotic consistency. Thus, we present here a general diagnostic procedure which directly and automatically evaluates the accuracy of the bootstrap's outputs, determining whether or not the bootstrap is performing satisfactorily when applied to a given dataset and estimator. We show that our proposed diagnostic is effective via an extensive empirical evaluation on a variety of estimators and simulated and real datasets, including a real-world query workload from Conviva, Inc. involving 1.7TB of data (i.e., approximately 0.5 billion data points).

This work introduces Divide-Factor-Combine (DFC), a parallel divide-and-conquer framework for noisy matrix factorization. DFC divides a large-scale matrix factorization task into smaller subproblems, solves each subproblem in parallel using an arbitrary base matrix factorization algorithm, and combines the subproblem solutions using techniques from randomized matrix approximation. Our experiments with collaborative filtering, video background modeling, and simulated data demonstrate the near-linear to super-linear speed-ups attainable with this approach. Moreover, our analysis shows that DFC enjoys high-probability recovery guarantees comparable to those of its base algorithm.

This paper addresses the problem of approximate singular value decomposition of large dense matrices that arises naturally in many machine learning applications. We discuss two recently introduced sampling-based spectral decomposition techniques: the Nyström and the Column-sampling methods. We present a theoretical comparison between the two methods and provide novel insights regarding their suitability for various applications. We then provide experimental results motivated by this theory. Finally, we propose an efficient adaptive sampling technique to select informative columns from the original matrix. This novel technique outperforms standard sampling methods on a variety of datasets.

A crucial technique for scaling kernel methods to very large data sets reaching or exceeding millions of instances is based on low-rank approximation of kernel matrices. We introduce a new family of algorithms based on mixtures of Nystrom approximations, ensemble Nystrom algorithms, that yield more accurate low-rank approximations than the standard Nystrom method. We give a detailed study of multiple variants of these algorithms based on simple averaging, an exponential weight method, or regression-based methods. We also present a theoretical analysis of these algorithms, including novel error bounds guaranteeing a better convergence rate than the standard Nystrom method. Finally, we report the results of extensive experiments with several data sets containing up to 1M points demonstrating the significant performance improvements gained over the standard Nystrom approximation.

This paper examines the problem of extracting low-dimensional manifold structure given millions of high-dimensional face images. Specifically, we address the computational challenges of nonlinear dimensionality reduction via Isomap and Laplacian Eigenmaps, using a graph containing about 18 million nodes and 65 million edges. Since most manifold learning techniques rely on spectral decomposition, we first analyze two approximate spectral decomposition techniques for large dense matrices (Nystrom and column-sampling), providing the first direct theoretical and empirical comparison between these techniques. We next show extensive experiments on learning low-dimensional embeddings for two large face datasets: CMU-PIE (35 thousand faces) and a web dataset (18 million faces). Our comparisons show that the Nystrom approximation is superior to the column-sampling method. Furthermore, approximate Isomap tends to perform better than Laplacian Eigenmaps on both clustering and classification with the labeled CMU-PIE dataset.

The problem of identifying the minimal gene set required to sustain life is of crucial importance in understanding cellular mechanisms and designing therapeutic drugs. This work describes several kernel-based solutions for predicting essential genes that outperform existing models while using less training data. Our first solution is based on a semi-manually designed kernel derived from the Pfam database, which includes several Pfam domains. We then present novel and general domain-based sequence kernels that capture sequence similarity with respect to several domains made of large sets of protein sequences. We show how to deal with the large size of the problem -- several thousands of domains with individual domains sometimes containing thousands of sequences -- by representing and efficiently computing these kernels using automata. We report results of extensive experiments demonstrating that they compare favorably with the Pfam kernel in predicting protein essentiality, while requiring no manual tuning.