Search

We conducted an experiment during the review process of the 2023 International Conference on Machine Learning (ICML) that requested authors with multiple submissions to rank their own papers based on perceived quality. We received 1,342 rankings, each from a distinct author, pertaining to 2,592 submissions. In this paper, we present an empirical analysis of how author-provided rankings could be leveraged to improve peer review processes at machine learning conferences. We focus on the Isotonic Mechanism, which calibrates raw review scores using author-provided rankings. Our analysis demonstrates that the ranking-calibrated scores outperform raw scores in estimating the ground truth ``expected review scores'' in both squared and absolute error metrics. Moreover, we propose several cautious, low-risk approaches to using the Isotonic Mechanism and author-provided rankings in peer review processes, including assisting senior area chairs' oversight of area chairs' recommendations, supporting the selection of paper awards, and guiding the recruitment of emergency reviewers. We conclude the paper by addressing the study's limitations and proposing future research directions., Comment: See more details about the experiment at https://openrank.cc/

Applied recommender systems research is in a curious position. While there is a very rigorous protocol for measuring performance by A/B testing, best practice for finding a `B' to test does not explicitly target performance but rather targets a proxy measure. The success or failure of a given A/B test then depends entirely on if the proposed proxy is better correlated to performance than the previous proxy. No principle exists to identify if one proxy is better than another offline, leaving the practitioners shooting in the dark. The purpose of this position paper is to question this anti-Utopian thinking and argue that a non-standard use of the deep learning stacks actually has the potential to unlock reward optimizing recommendation., Comment: 11 pages

Compositional generalization is one of the main properties which differentiates lexical learning in humans from state-of-art neural networks. We propose a general framework for building models that can generalize compositionally using the concept of Generalized Grammar Rules (GGRs), a class of symmetry-based compositional constraints for transduction tasks, which we view as a transduction analogue of equivariance constraints in physics-inspired tasks. Besides formalizing generalized notions of symmetry for language transduction, our framework is general enough to contain many existing works as special cases. We present ideas on how GGRs might be implemented, and in the process draw connections to reinforcement learning and other areas of research., Comment: 12 pages

In the past decade, advances in deep learning have resulted in breakthroughs in a variety of areas, including computer vision, natural language understanding, speech recognition, and reinforcement learning. Specialized, high-performing neural architectures are crucial to the success of deep learning in these areas. Neural architecture search (NAS), the process of automating the design of neural architectures for a given task, is an inevitable next step in automating machine learning and has already outpaced the best human-designed architectures on many tasks. In the past few years, research in NAS has been progressing rapidly, with over 1000 papers released since 2020 (Deng and Lindauer, 2021). In this survey, we provide an organized and comprehensive guide to neural architecture search. We give a taxonomy of search spaces, algorithms, and speedup techniques, and we discuss resources such as benchmarks, best practices, other surveys, and open-source libraries.

This is a collection of (mostly) pen-and-paper exercises in machine learning. The exercises are on the following topics: linear algebra, optimisation, directed graphical models, undirected graphical models, expressive power of graphical models, factor graphs and message passing, inference for hidden Markov models, model-based learning (including ICA and unnormalised models), sampling and Monte-Carlo integration, and variational inference., Comment: The associated github page is https://github.com/michaelgutmann/ml-pen-and-paper-exercises

I consider a setting where reviewers offer very noisy scores for several items for the selection of high-quality ones (e.g., peer review of large conference proceedings), whereas the owner of these items knows the true underlying scores but prefers not to provide this information. To address this withholding of information, in this paper, I introduce the Isotonic Mechanism, a simple and efficient approach to improving imprecise raw scores by leveraging certain information that the owner is incentivized to provide. This mechanism takes the ranking of the items from best to worst provided by the owner as input, in addition to the raw scores provided by the reviewers. It reports the adjusted scores for the items by solving a convex optimization problem. Under certain conditions, I show that the owner's optimal strategy is to honestly report the true ranking of the items to her best knowledge in order to maximize the expected utility. Moreover, I prove that the adjusted scores provided by this owner-assisted mechanism are significantly more accurate than the raw scores provided by the reviewers. This paper concludes with several extensions of the Isotonic Mechanism and some refinements of the mechanism for practical consideration., Comment: Corrected typos and added a reference

Complex chemical structures, like drugs, are usually defined by SMILES strings as a sequence of molecules and bonds. These SMILES strings are used in different complex machine learning-based drug-related research and representation works. Escaping from complex representation, in this work, we pose a single question: What if we treat drug SMILES as conventional sentences and engage in text classification for drug classification? Our experiments affirm the possibility with very competitive scores. The study explores the notion of viewing each atom and bond as sentence components, employing basic NLP methods to categorize drug types, proving that complex problems can also be solved with simpler perspectives. The data and code are available here: https://github.com/azminewasi/Drug-Classification-NLP., Comment: 7 pages, 2 figures, 5 tables, Accepted (invited to present) to the The Second Tiny Papers Track at ICLR 2024 (https://openreview.net/forum?id=VUYCyH8fCw)

This paper illustrates the central role of loss functions in data-driven decision making, providing a comprehensive survey on their influence in cost-sensitive classification (CSC) and reinforcement learning (RL). We demonstrate how different regression loss functions affect the sample efficiency and adaptivity of value-based decision making algorithms. Across multiple settings, we prove that algorithms using the binary cross-entropy loss achieve first-order bounds scaling with the optimal policy's cost and are much more efficient than the commonly used squared loss. Moreover, we prove that distributional algorithms using the maximum likelihood loss achieve second-order bounds scaling with the policy variance and are even sharper than first-order bounds. This in particular proves the benefits of distributional RL. We hope that this paper serves as a guide analyzing decision making algorithms with varying loss functions, and can inspire the reader to seek out better loss functions to improve any decision making algorithm.

Uncertainty quantification plays a key role in the development of autonomous systems, decision-making, and tracking over wireless sensor networks (WSNs). However, there is a need of providing uncertainty confidence bounds, especially for distributed machine learning-based tracking, dealing with different volumes of data collected by sensors. This paper aims to fill in this gap and proposes a distributed Gaussian process (DGP) approach for point target tracking and derives upper confidence bounds (UCBs) of the state estimates. A unique contribution of this paper includes the derived theoretical guarantees on the proposed approach and its maximum accuracy for tracking with and without clutter measurements. Particularly, the developed approaches with uncertainty bounds are generic and can provide trustworthy solutions with an increased level of reliability. A novel hybrid Bayesian filtering method is proposed to enhance the DGP approach by adopting a Poisson measurement likelihood model. The proposed approaches are validated over a WSN case study, where sensors have limited sensing ranges. Numerical results demonstrate the tracking accuracy and robustness of the proposed approaches. The derived UCBs constitute a tool for trustworthiness evaluation of DGP approaches. The simulation results reveal that the proposed UCBs successfully encompass the true target states with 88% and 42% higher probability in X and Y coordinates, respectively, when compared to the confidence interval-based method.

Variational inference (VI) is a computationally efficient and scalable methodology for approximate Bayesian inference. It strikes a balance between accuracy of uncertainty quantification and practical tractability. It excels at generative modelling and inversion tasks due to its built-in Bayesian regularisation and flexibility, essential qualities for physics related problems. Deriving the central learning objective for VI must often be tailored to new learning tasks where the nature of the problems dictates the conditional dependence between variables of interest, such as arising in physics problems. In this paper, we provide an accessible and thorough technical introduction to VI for forward and inverse problems, guiding the reader through standard derivations of the VI framework and how it can best be realized through deep learning. We then review and unify recent literature exemplifying the creative flexibility allowed by VI. This paper is designed for a general scientific audience looking to solve physics-based problems with an emphasis on uncertainty quantification.

In today's era, Neural Networks (NN) are applied in various scientific fields such as robotics, medicine, engineering, etc. However, the predictions of neural networks themselves contain a degree of uncertainty that must always be taken into account before any decision is made. This is why many researchers have focused on developing different ways to quantify the uncertainty of neural network predictions. Some of these methods are based on generating prediction intervals (PI) via neural networks for the requested target values. The SEF (Shifting the Error Function) method presented in this paper is a new method that belongs to this category of methods. The proposed approach involves training a single neural network three times, thus generating an estimate along with the corresponding upper and lower bounds for a given problem. A pivotal aspect of the method is the calculation of a parameter from the initial network's estimates, which is then integrated into the loss functions of the other two networks. This innovative process effectively produces PIs, resulting in a robust and efficient technique for uncertainty quantification. To evaluate the effectiveness of our method, a comparison in terms of successful PI generation between the SEF, PI3NN and PIVEN methods was made using two synthetic datasets., Comment: The paper has been accepted at the 2024 International Conference on Computer and Applications (ICCA24), Cairo, Egypt, December 17-19, 2024. https://icca-conf.info/icca-2024

This paper considers the scenario in which there are multiple institutions, each with a limited capacity for candidates, and candidates, each with preferences over the institutions. A central entity evaluates the utility of each candidate to the institutions, and the goal is to select candidates for each institution in a way that maximizes utility while also considering the candidates' preferences. The paper focuses on the setting in which candidates are divided into multiple groups and the observed utilities of candidates in some groups are biased--systematically lower than their true utilities. The first result is that, in these biased settings, prior algorithms can lead to selections with sub-optimal true utility and significant discrepancies in the fraction of candidates from each group that get their preferred choices. Subsequently, an algorithm is presented along with proof that it produces selections that achieve near-optimal group fairness with respect to preferences while also nearly maximizing the true utility under distributional assumptions. Further, extensive empirical validation of these results in real-world and synthetic settings, in which the distributional assumptions may not hold, are presented., Comment: The conference version of this paper appears in ICML 2024

In these notes, we prove a recent conjecture posed in the paper by R\"ais\"a, O. et al. [Subsampling is not Magic: Why Large Batch Sizes Work for Differentially Private Stochastic Optimization (2024)]. Theorem 6.2 of the paper asserts that for the Sampled Gaussian Mechanism - a composition of subsampling and additive Gaussian noise, the effective noise level, $\sigma_{\text{eff}} = \frac{\sigma(q)}{q}$, decreases as a function of the subsampling rate $q$. Consequently, larger subsampling rates are preferred for better privacy-utility trade-offs. Our notes provide a rigorous proof of Conjecture 6.3, which was left unresolved in the original paper, thereby completing the proof of Theorem 6.2.

In this brief paper, we present a naive aggregation algorithm for a typical learning problem with expert advice setting, in which the task of improving generalization, i.e., model validation, is embedded in the learning process as a sequential decision-making problem. In particular, we consider a class of learning problem of point estimations for modeling high-dimensional nonlinear functions, where a group of experts update their parameter estimates using the discrete-time version of gradient systems, with small additive noise term, guided by the corresponding subsample datasets obtained from the original dataset. Here, our main objective is to provide conditions under which such an algorithm will sequentially determine a set of mixing distribution strategies used for aggregating the experts' estimates that ultimately leading to an optimal parameter estimate, i.e., as a consensus solution for all experts, which is better than any individual expert's estimate in terms of improved generalization or learning performances. Finally, as part of this work, we present some numerical results for a typical case of nonlinear regression problem., Comment: Brief paper, with 7 pages, 1 figure

We propose a general approach for encouraging fairness in survival analysis models based on minimizing a worst-case error across all subpopulations that occur with at least a user-specified probability. This approach can be used to convert many existing survival analysis models into ones that simultaneously encourage fairness, without requiring the user to specify which attributes or features to treat as sensitive in the training loss function. From a technical standpoint, our approach applies recent developments of distributionally robust optimization (DRO) to survival analysis. The complication is that existing DRO theory uses a training loss function that decomposes across contributions of individual data points, i.e., any term that shows up in the loss function depends only on a single training point. This decomposition does not hold for commonly used survival loss functions, including for the Cox proportional hazards model, its deep neural network variants, and many other recently developed models that use loss functions involving ranking or similarity score calculations. We address this technical hurdle using a sample splitting strategy. We demonstrate our sample splitting DRO approach by using it to create fair versions of a diverse set of existing survival analysis models including the Cox model (and its deep variant DeepSurv), the discrete-time model DeepHit, and the neural ODE model SODEN. We also establish a finite-sample theoretical guarantee to show what our sample splitting DRO loss converges to. For the Cox model, we further derive an exact DRO approach that does not use sample splitting. For all the models that we convert into DRO variants, we show that the DRO variants often score better on recently established fairness metrics (without incurring a significant drop in accuracy) compared to existing survival analysis fairness regularization techniques., Comment: Accepted at the Journal of Machine Learning Research; this paper is a journal paper extension of our earlier Machine Learning for Health 2022 paper (arXiv:2211.10508)

Renewable prosumers face the complex challenge of balancing self-sufficiency with seamless grid and market integration. This paper introduces a novel prosumers network framework aimed at mitigating peak loads in Iran, particularly under the uncertainties inherent in renewable energy generation and demand. A cost-oriented integrated prediction and optimization approach is proposed, empowering prosumers to make informed decisions within a distributed contextual stochastic optimization (DCSO) framework. The problem is formulated as a bi-level two-stage multi-time scale optimization to determine optimal operation and interaction strategies under various scenarios, considering flexible resources. To facilitate grid integration, a novel consensus-based contextual information sharing mechanism is proposed. This approach enables coordinated collective behaviors and leverages contextual data more effectively. The overall problem is recast as a mixed-integer linear program (MILP) by incorporating optimality conditions and linearizing complementarity constraints. Additionally, a distributed algorithm using the consensus alternating direction method of multipliers (ADMM) is presented for computational tractability and privacy preservation. Numerical results highlights that integrating prediction with optimization and implementing a contextual information-sharing network among prosumers significantly reduces peak loads as well as total costs., Comment: 10 pages, 26 figure, journal paper

In this paper, we study the statistical and geometrical properties of the Kullback-Leibler divergence with kernel covariance operators (KKL) introduced by Bach [2022]. Unlike the classical Kullback-Leibler (KL) divergence that involves density ratios, the KKL compares probability distributions through covariance operators (embeddings) in a reproducible kernel Hilbert space (RKHS), and compute the Kullback-Leibler quantum divergence. This novel divergence hence shares parallel but different aspects with both the standard Kullback-Leibler between probability distributions and kernel embeddings metrics such as the maximum mean discrepancy. A limitation faced with the original KKL divergence is its inability to be defined for distributions with disjoint supports. To solve this problem, we propose in this paper a regularised variant that guarantees that the divergence is well defined for all distributions. We derive bounds that quantify the deviation of the regularised KKL to the original one, as well as finite-sample bounds. In addition, we provide a closed-form expression for the regularised KKL, specifically applicable when the distributions consist of finite sets of points, which makes it implementable. Furthermore, we derive a Wasserstein gradient descent scheme of the KKL divergence in the case of discrete distributions, and study empirically its properties to transport a set of points to a target distribution.

Feature selection is a crucial technique for handling high-dimensional data. In unsupervised scenarios, many popular algorithms focus on preserving the original data structure. In this paper, we reproduce the IVFS algorithm introduced in AAAI 2020, which is inspired by the random subset method and preserves data similarity by maintaining topological structure. We systematically organize the mathematical foundations of IVFS and validate its effectiveness through numerical experiments similar to those in the original paper. The results demonstrate that IVFS outperforms SPEC and MCFS on most datasets, although issues with its convergence and stability persist., Comment: arXiv admin note: substantial text overlap with arXiv:2004.01299 by other authors

Shapley values have seen widespread use in machine learning as a way to explain model predictions and estimate the importance of covariates. Accurately explaining models is critical in real-world models to both aid in decision making and to infer the properties of the true data-generating process (DGP). In this paper, we demonstrate that while model-based Shapley values might be accurate explainers of model predictions, machine learning models themselves are often poor explainers of the DGP even if the model is highly accurate. Particularly in the presence of interrelated or noisy variables, the output of a highly predictive model may fail to account for these relationships. This implies explanations of a trained model's behavior may fail to provide meaningful insight into the DGP. In this paper we introduce a novel variable importance algorithm, Shapley Marginal Surplus for Strong Models, that samples the space of possible models to come up with an inferential measure of feature importance. We compare this method to other popular feature importance methods, both Shapley-based and non-Shapley based, and demonstrate significant outperformance in inferential capabilities relative to other methods.

Representation learning is an approach that allows to discover and extract the factors of variation from the data. Intuitively, a representation is said to be disentangled if it separates the different factors of variation in a way that is understandable to humans. Definitions of disentanglement and metrics to measure it usually assume that the factors of variation are independent of each other. However, this is generally false in the real world, which limits the use of these definitions and metrics to very specific and unrealistic scenarios. In this paper we give a definition of disentanglement based on information theory that is also valid when the factors of variation are not independent. Furthermore, we relate this definition to the Information Bottleneck Method. Finally, we propose a method to measure the degree of disentanglement from the given definition that works when the factors of variation are not independent. We show through different experiments that the method proposed in this paper correctly measures disentanglement with non-independent factors of variation, while other methods fail in this scenario.

Supervised operator learning centers on the use of training data, in the form of input-output pairs, to estimate maps between infinite-dimensional spaces. It is emerging as a powerful tool to complement traditional scientific computing, which may often be framed in terms of operators mapping between spaces of functions. Building on the classical random features methodology for scalar regression, this paper introduces the function-valued random features method. This leads to a supervised operator learning architecture that is practical for nonlinear problems yet is structured enough to facilitate efficient training through the optimization of a convex, quadratic cost. Due to the quadratic structure, the trained model is equipped with convergence guarantees and error and complexity bounds, properties that are not readily available for most other operator learning architectures. At its core, the proposed approach builds a linear combination of random operators. This turns out to be a low-rank approximation of an operator-valued kernel ridge regression algorithm, and hence the method also has strong connections to Gaussian process regression. The paper designs function-valued random features that are tailored to the structure of two nonlinear operator learning benchmark problems arising from parametric partial differential equations. Numerical results demonstrate the scalability, discretization invariance, and transferability of the function-valued random features method., Comment: 36 pages, 1 table, 9 figures. SIGEST version of SIAM J. Sci. Comput. Vol. 43 No. 5 (2021) pp. A3212-A3243, hence text overlap with arXiv:2005.10224

This paper studies Kernel density estimation for a high-dimensional distribution $\rho(x)$. Traditional approaches have focused on the limit of large number of data points $n$ and fixed dimension $d$. We analyze instead the regime where both the number $n$ of data points $y_i$ and their dimensionality $d$ grow with a fixed ratio $\alpha=(\log n)/d$. Our study reveals three distinct statistical regimes for the kernel-based estimate of the density $\hat \rho_h^{\mathcal {D}}(x)=\frac{1}{n h^d}\sum_{i=1}^n K\left(\frac{x-y_i}{h}\right)$, depending on the bandwidth $h$: a classical regime for large bandwidth where the Central Limit Theorem (CLT) holds, which is akin to the one found in traditional approaches. Below a certain value of the bandwidth, $h_{CLT}(\alpha)$, we find that the CLT breaks down. The statistics of $\hat \rho_h^{\mathcal {D}}(x)$ for a fixed $x$ drawn from $\rho(x)$ is given by a heavy-tailed distribution (an alpha-stable distribution). In particular below a value $h_G(\alpha)$, we find that $\hat \rho_h^{\mathcal {D}}(x)$ is governed by extreme value statistics: only a few points in the database matter and give the dominant contribution to the density estimator. We provide a detailed analysis for high-dimensional multivariate Gaussian data. We show that the optimal bandwidth threshold based on Kullback-Leibler divergence lies in the new statistical regime identified in this paper. Our findings reveal limitations of classical approaches, show the relevance of these new statistical regimes, and offer new insights for Kernel density estimation in high-dimensional settings.

Attributed graph clustering, which aims to group the nodes of an attributed graph into disjoint clusters, has made promising advancements in recent years. However, most existing methods face challenges when applied to large graphs due to the expensive computational cost and high memory usage. In this paper, we introduce Scalable and Adaptive Spectral Embedding (SASE), a simple attributed graph clustering method devoid of parameter learning. SASE comprises three main components: node features smoothing via $k$-order simple graph convolution, scalable spectral clustering using random Fourier features, and adaptive order selection. With these designs, SASE not only effectively captures global cluster structures but also exhibits linear time and space complexity relative to the graph size. Empirical results demonstrate the superiority of SASE. For example, on the ArXiv dataset with 169K nodes and 1.17M edges, SASE achieves a 6.9\% improvement in ACC and a $5.87\times$ speedup compared to the runner-up, S3GC., Comment: Accepted by CIKM 2024 (Short Paper)

With low-cost computing devices, improved sensor technology, and the proliferation of data-driven algorithms, we have more data than we know what to do with. In transportation, we are seeing a surge in spatiotemporal data collection. At the same time, concerns over user privacy have led to research on differential privacy in applied settings. In this paper, we look at some recent developments in differential privacy in the context of spatiotemporal data. Spatiotemporal data contain not only features about users but also the geographical locations of their frequent visits. Hence, the public release of such data carries extreme risks. To address the need for such data in research and inference without exposing private information, significant work has been proposed. This survey paper aims to summarize these efforts and provide a review of differential privacy mechanisms and related software. We also discuss related work in transportation where such mechanisms have been applied. Furthermore, we address the challenges in the deployment and mass adoption of differential privacy in transportation spatiotemporal data for downstream analyses.

Plant phenotyping is the assessment of a plant's traits and plant identification is the process of determining the category such as genus and species. In this paper we present an interpretable neural network trained on the UPWINS spectral library which contains spectra with rich metadata across variation in species, health, growth stage, annual variation, and environmental conditions for 13 selected indicator species and natural common background species. We show that the neurons in the network learn spectral indicators for chemical and physiological traits through visualization of the network weights, and we show how these traits are combined by the network for species identification with an accuracy around 90% on a test set. While neural networks are often perceived as `black box' classifiers, our work shows that they can be in fact more explainable and informative than other machine learning methods. We show that the neurons learn fundamental traits about the vegetation, for example the composition of different types of chlorophyll present which indicates species as well as response to illumination conditions. There is clear excess training capacity in our network, and we expect that as the UPWINS spectral library continues to grow the approach in this paper will provide further foundational insights in understanding plant traits. This provides a methodology for designing and interpreting neural networks on spectral data in general, and provides a framework for using neural networks with hyperspectral imagery for understanding vegetation that is extendable to other domains.

[See paper for full abstract] Meta-analysis is a crucial tool for answering scientific questions. It is usually conducted on a relatively small amount of ``trusted'' data -- ideally from randomized, controlled trials -- which allow causal effects to be reliably estimated with minimal assumptions. We show how to answer causal questions much more precisely by making two changes. First, we incorporate untrusted data drawn from large observational databases, related scientific literature and practical experience -- without sacrificing rigor or introducing strong assumptions. Second, we train richer models capable of handling heterogeneous trials, addressing a long-standing challenge in meta-analysis. Our approach is based on conformal prediction, which fundamentally produces rigorous prediction intervals, but doesn't handle indirect observations: in meta-analysis, we observe only noisy effects due to the limited number of participants in each trial. To handle noise, we develop a simple, efficient version of fully-conformal kernel ridge regression, based on a novel condition called idiocentricity. We introduce noise-correcting terms in the residuals and analyze their interaction with a ``variance shaving'' technique. In multiple experiments on healthcare datasets, our algorithms deliver tighter, sounder intervals than traditional ones. This paper charts a new course for meta-analysis and evidence-based medicine, where heterogeneity and untrusted data are embraced for more nuanced and precise predictions., Comment: Full-length version of conference submission

Graph clustering, which aims to divide a graph into several homogeneous groups, is a critical area of study with applications that span various fields such as social network analysis, bioinformatics, and image segmentation. This paper explores both traditional and more recent approaches to graph clustering. Firstly, key concepts and definitions in graph theory are introduced. The background section covers essential topics, including graph Laplacians and the integration of Deep Learning in graph analysis. The paper then delves into traditional clustering methods, including Spectral Clustering and the Leiden algorithm. Following this, state-of-the-art clustering techniques that leverage deep learning are examined. A comprehensive comparison of these methods is made through experiments. The paper concludes with a discussion of the practical applications of graph clustering and potential future research directions.

Hyperspherical Prototypical Learning (HPL) is a supervised approach to representation learning that designs class prototypes on the unit hypersphere. The prototypes bias the representations to class separation in a scale invariant and known geometry. Previous approaches to HPL have either of the following shortcomings: (i) they follow an unprincipled optimisation procedure; or (ii) they are theoretically sound, but are constrained to only one possible latent dimension. In this paper, we address both shortcomings. To address (i), we present a principled optimisation procedure whose solution we show is optimal. To address (ii), we construct well-separated prototypes in a wide range of dimensions using linear block codes. Additionally, we give a full characterisation of the optimal prototype placement in terms of achievable and converse bounds, showing that our proposed methods are near-optimal., Comment: 14 pages: 9 of the main paper, 2 of references, and 3 of appendices. To appear in the Proceedings of the Geometry-grounded Representation Learning and Generative Modeling at the 41st International Conference on Machine Learning, Vienna, Austria. Code is available at: https://github.com/martinlindstrom/coding_theoretic_hpl

Since the seminal paper of Hendrycks et al. arXiv:1610.02136, Post-hoc deep Out-of-Distribution (OOD) detection has expanded rapidly. As a result, practitioners working on safety-critical applications and seeking to improve the robustness of a neural network now have a plethora of methods to choose from. However, no method outperforms every other on every dataset arXiv:2210.07242, so the current best practice is to test all the methods on the datasets at hand. This paper shifts focus from developing new methods to effectively combining existing ones to enhance OOD detection. We propose and compare four different strategies for integrating multiple detection scores into a unified OOD detector, based on techniques such as majority vote, empirical and copulas-based Cumulative Distribution Function modeling, and multivariate quantiles based on optimal transport. We extend common OOD evaluation metrics -- like AUROC and FPR at fixed TPR rates -- to these multi-dimensional OOD detectors, allowing us to evaluate them and compare them with individual methods on extensive benchmarks. Furthermore, we propose a series of guidelines to choose what OOD detectors to combine in more realistic settings, i.e. in the absence of known OOD data, relying on principles drawn from Outlier Exposure arXiv:1812.04606. The code is available at https://github.com/paulnovello/multi-ood.

This paper investigates the problem of regression model generation. A model is a superposition of primitive functions. The model structure is described by a weighted colored graph. Each graph vertex corresponds to some primitive function. An edge assigns a superposition of two functions. The weight of an edge equals the probability of superposition. To generate an optimal model one has to reconstruct its structure from its graph adjacency matrix. The proposed algorithm reconstructs the~minimum spanning tree from the~weighted colored graph. This paper presents a novel solution based on the prize-collecting Steiner tree algorithm. This algorithm is compared with its alternatives.

In this paper, we explore advanced modifications to the Tweedie regression model in order to address its limitations in modeling aggregate claims for various types of insurance such as automobile, health, and liability. Traditional Tweedie models, while effective in capturing the probability and magnitude of claims, usually fall short in accurately representing the large incidence of zero claims. Our recommended approach involves a refined modeling of the zero-claim process, together with the integration of boosting methods in order to help leverage an iterative process to enhance predictive accuracy. Despite the inherent slowdown in learning algorithms due to this iteration, several efficient implementation techniques that also help precise tuning of parameter like XGBoost, LightGBM, and CatBoost have emerged. Nonetheless, we chose to utilize CatBoost, a efficient boosting approach that effectively handles categorical and other special types of data. The core contribution of our paper is the assembly of separate modeling for zero claims and the application of tree-based boosting ensemble methods within a CatBoost framework, assuming that the inflated probability of zero is a function of the mean parameter. The efficacy of our enhanced Tweedie model is demonstrated through the application of an insurance telematics dataset, which presents the additional complexity of compositional feature variables. Our modeling results reveal a marked improvement in model performance, showcasing its potential to deliver more accurate predictions suitable for insurance claim analytics.

Feature selection in Knowledge Graphs (KGs) are increasingly utilized in diverse domains, including biomedical research, Natural Language Processing (NLP), and personalized recommendation systems. This paper delves into the methodologies for feature selection within KGs, emphasizing their roles in enhancing machine learning (ML) model efficacy, hypothesis generation, and interpretability. Through this comprehensive review, we aim to catalyze further innovation in feature selection for KGs, paving the way for more insightful, efficient, and interpretable analytical models across various domains. Our exploration reveals the critical importance of scalability, accuracy, and interpretability in feature selection techniques, advocating for the integration of domain knowledge to refine the selection process. We highlight the burgeoning potential of multi-objective optimization and interdisciplinary collaboration in advancing KG feature selection, underscoring the transformative impact of such methodologies on precision medicine, among other fields. The paper concludes by charting future directions, including the development of scalable, dynamic feature selection algorithms and the integration of explainable AI principles to foster transparency and trust in KG-driven models.

Explainable Artificial Intelligence (XAI) methods help to understand the internal mechanism of machine learning models and how they reach a specific decision or made a specific action. The list of informative features is one of the most common output of XAI methods. Multicollinearity is one of the big issue that should be considered when XAI generates the explanation in terms of the most informative features in an AI system. No review has been dedicated to investigate the current approaches to handle such significant issue. In this paper, we provide a review of the current state-of-the-art approaches in relation to the XAI in the context of recent advances in dealing with the multicollinearity issue. To do so, we searched in three repositories that are: Web of Science, Scopus and IEEE Xplore to find pertinent published papers. After excluding irrelevant papers, seven papers were considered in the review. In addition, we discuss the current XAI methods and their limitations in dealing with the multicollinearity and suggest future directions.

Learning a parametric model from a given dataset indeed enables to capture intrinsic dependencies between random variables via a parametric conditional probability distribution and in turn predict the value of a label variable given observed variables. In this paper, we undertake a comparative analysis of generative and discriminative approaches which differ in their construction and the structure of the underlying inference problem. Our objective is to compare the ability of both approaches to leverage information from various sources in an epistemic uncertainty aware inference via the posterior predictive distribution. We assess the role of a prior distribution, explicit in the generative case and implicit in the discriminative case, leading to a discussion about discriminative models suffering from imbalanced dataset. We next examine the double role played by the observed variables in the generative case, and discuss the compatibility of both approaches with semi-supervised learning. We also provide with practical insights and we examine how the modeling choice impacts the sampling from the posterior predictive distribution. With regard to this, we propose a general sampling scheme enabling supervised learning for both approaches, as well as semi-supervised learning when compatible with the considered modeling approach. Throughout this paper, we illustrate our arguments and conclusions using the example of affine regression, and validate our comparative analysis through classification simulations using neural network based models.

Causal interactions among a group of variables are often modeled by a single causal graph. In some domains, however, these interactions are best described by multiple co-existing causal graphs, e.g., in dynamical systems or genomics. This paper addresses the hitherto unknown role of interventions in learning causal interactions among variables governed by a mixture of causal systems, each modeled by one directed acyclic graph (DAG). Causal discovery from mixtures is fundamentally more challenging than single-DAG causal discovery. Two major difficulties stem from (i) inherent uncertainty about the skeletons of the component DAGs that constitute the mixture and (ii) possibly cyclic relationships across these component DAGs. This paper addresses these challenges and aims to identify edges that exist in at least one component DAG of the mixture, referred to as true edges. First, it establishes matching necessary and sufficient conditions on the size of interventions required to identify the true edges. Next, guided by the necessity results, an adaptive algorithm is designed that learns all true edges using ${\cal O}(n^2)$ interventions, where $n$ is the number of nodes. Remarkably, the size of the interventions is optimal if the underlying mixture model does not contain cycles across its components. More generally, the gap between the intervention size used by the algorithm and the optimal size is quantified. It is shown to be bounded by the cyclic complexity number of the mixture model, defined as the size of the minimal intervention that can break the cycles in the mixture, which is upper bounded by the number of cycles among the ancestors of a node.

Training Deep Neural Networks (DNNs) with adversarial examples often results in poor generalization to test-time adversarial data. This paper investigates this issue, known as adversarially robust generalization, through the lens of Rademacher complexity. Building upon the studies by Khim and Loh (2018); Yin et al. (2019), numerous works have been dedicated to this problem, yet achieving a satisfactory bound remains an elusive goal. Existing works on DNNs either apply to a surrogate loss instead of the robust loss or yield bounds that are notably looser compared to their standard counterparts. In the latter case, the bounds have a higher dependency on the width $m$ of the DNNs or the dimension $d$ of the data, with an extra factor of at least $\mathcal{O}(\sqrt{m})$ or $\mathcal{O}(\sqrt{d})$. This paper presents upper bounds for adversarial Rademacher complexity of DNNs that match the best-known upper bounds in standard settings, as established in the work of Bartlett et al. (2017), with the dependency on width and dimension being $\mathcal{O}(\ln(dm))$. The central challenge addressed is calculating the covering number of adversarial function classes. We aim to construct a new cover that possesses two properties: 1) compatibility with adversarial examples, and 2) precision comparable to covers used in standard settings. To this end, we introduce a new variant of covering number called the \emph{uniform covering number}, specifically designed and proven to reconcile these two properties. Consequently, our method effectively bridges the gap between Rademacher complexity in robust and standard generalization., Comment: COLT 2024

This paper introduces a novel information-theoretic perspective on the relationship between prominent group fairness notions in machine learning, namely statistical parity, equalized odds, and predictive parity. It is well known that simultaneous satisfiability of these three fairness notions is usually impossible, motivating practitioners to resort to approximate fairness solutions rather than stringent satisfiability of these definitions. However, a comprehensive analysis of their interrelations, particularly when they are not exactly satisfied, remains largely unexplored. Our main contribution lies in elucidating an exact relationship between these three measures of (un)fairness by leveraging a body of work in information theory called partial information decomposition (PID). In this work, we leverage PID to identify the granular regions where these three measures of (un)fairness overlap and where they disagree with each other leading to potential tradeoffs. We also include numerical simulations to complement our results., Comment: Published as a conference paper at 2024 IEEE International Symposium on Information Theory (ISIT 2024)

This paper explores the use of XGBoost for composite quantile regression. XGBoost is a highly popular model renowned for its flexibility, efficiency, and capability to deal with missing data. The optimization uses a second order approximation of the loss function, complicating the use of loss functions with a zero or vanishing second derivative. Quantile regression -- a popular approach to obtain conditional quantiles when point estimates alone are insufficient -- unfortunately uses such a loss function, the pinball loss. Existing workarounds are typically inefficient and can result in severe quantile crossings. In this paper, we present a smooth approximation of the pinball loss, the arctan pinball loss, that is tailored to the needs of XGBoost. Specifically, contrary to other smooth approximations, the arctan pinball loss has a relatively large second derivative, which makes it more suitable to use in the second order approximation. Using this loss function enables the simultaneous prediction of multiple quantiles, which is more efficient and results in far fewer quantile crossings., Comment: 24 pages, 9 figures

Actor-critic (AC) is a powerful method for learning an optimal policy in reinforcement learning, where the critic uses algorithms, e.g., temporal difference (TD) learning with function approximation, to evaluate the current policy and the actor updates the policy along an approximate gradient direction using information from the critic. This paper provides the \textit{tightest} non-asymptotic convergence bounds for both the AC and natural AC (NAC) algorithms. Specifically, existing studies show that AC converges to an $\epsilon+\varepsilon_{\text{critic}}$ neighborhood of stationary points with the best known sample complexity of $\mathcal{O}(\epsilon^{-2})$ (up to a log factor), and NAC converges to an $\epsilon+\varepsilon_{\text{critic}}+\sqrt{\varepsilon_{\text{actor}}}$ neighborhood of the global optimum with the best known sample complexity of $\mathcal{O}(\epsilon^{-3})$, where $\varepsilon_{\text{critic}}$ is the approximation error of the critic and $\varepsilon_{\text{actor}}$ is the approximation error induced by the insufficient expressive power of the parameterized policy class. This paper analyzes the convergence of both AC and NAC algorithms with compatible function approximation. Our analysis eliminates the term $\varepsilon_{\text{critic}}$ from the error bounds while still achieving the best known sample complexities. Moreover, we focus on the challenging single-loop setting with a single Markovian sample trajectory. Our major technical novelty lies in analyzing the stochastic bias due to policy-dependent and time-varying compatible function approximation in the critic, and handling the non-ergodicity of the MDP due to the single Markovian sample trajectory. Numerical results are also provided in the appendix., Comment: ICML 2024

The adaptation and use of Machine Learning (ML) in our daily lives has led to concerns in lack of transparency, privacy, reliability, among others. As a result, we are seeing research in niche areas such as interpretability, causality, bias and fairness, and reliability. In this survey paper, we focus on a critical concern for adaptation of ML in risk-sensitive applications, namely understanding and quantifying uncertainty. Our paper approaches this topic in a structured way, providing a review of the literature in the various facets that uncertainty is enveloped in the ML process. We begin by defining uncertainty and its categories (e.g., aleatoric and epistemic), understanding sources of uncertainty (e.g., data and model), and how uncertainty can be assessed in terms of uncertainty quantification techniques (Ensembles, Bayesian Neural Networks, etc.). As part of our assessment and understanding of uncertainty in the ML realm, we cover metrics for uncertainty quantification for a single sample, dataset, and metrics for accuracy of the uncertainty estimation itself. This is followed by discussions on calibration (model and uncertainty), and decision making under uncertainty. Thus, we provide a more complete treatment of uncertainty: from the sources of uncertainty to the decision-making process. We have focused the review of uncertainty quantification methods on Deep Learning (DL), while providing the necessary background for uncertainty discussion within ML in general. Key contributions in this review are broadening the scope of uncertainty discussion, as well as an updated review of uncertainty quantification methods in DL.

Mathematical equivalence between statistical mechanics and machine learning theory has been known since the 20th century, and researches based on such equivalence have provided novel methodology in both theoretical physics and statistical learning theory. For example, algebraic approach in statistical mechanics such as operator algebra enables us to analyze phase transition phenomena mathematically. In this paper, for theoretical physicists who are interested in artificial intelligence, we review and prospect algebraic researches in machine learning theory. If a learning machine has hierarchical structure or latent variables, then the random Hamiltonian cannot be expressed by any quadratic perturbation because it has singularities. To study an equilibrium state defined by such a singular random Hamiltonian, algebraic approach is necessary to derive asymptotic form of the free energy and the generalization error. We also introduce the most recent advance, in fact, theoretical foundation for alignment of artificial intelligence is now being constructed based on algebraic learning theory. This paper is devoted to the memory of Professor Huzihiro Araki who is a pioneer founder of algebraic research in both statistical mechanics and quantum field theory.

This paper studies the challenging black-box adversarial attack that aims to generate adversarial examples against a black-box model by only using output feedback of the model to input queries. Some previous methods improve the query efficiency by incorporating the gradient of a surrogate white-box model into query-based attacks due to the adversarial transferability. However, the localized gradient is not informative enough, making these methods still query-intensive. In this paper, we propose a Prior-guided Bayesian Optimization (P-BO) algorithm that leverages the surrogate model as a global function prior in black-box adversarial attacks. As the surrogate model contains rich prior information of the black-box one, P-BO models the attack objective with a Gaussian process whose mean function is initialized as the surrogate model's loss. Our theoretical analysis on the regret bound indicates that the performance of P-BO may be affected by a bad prior. Therefore, we further propose an adaptive integration strategy to automatically adjust a coefficient on the function prior by minimizing the regret bound. Extensive experiments on image classifiers and large vision-language models demonstrate the superiority of the proposed algorithm in reducing queries and improving attack success rates compared with the state-of-the-art black-box attacks. Code is available at https://github.com/yibo-miao/PBO-Attack., Comment: ICML 2024

We study the trajectory of iterations and the convergence rates of the Expectation-Maximization (EM) algorithm for two-component Mixed Linear Regression (2MLR). The fundamental goal of MLR is to learn the regression models from unlabeled observations. The EM algorithm finds extensive applications in solving the mixture of linear regressions. Recent results have established the super-linear convergence of EM for 2MLR in the noiseless and high SNR settings under some assumptions and its global convergence rate with random initialization has been affirmed. However, the exponent of convergence has not been theoretically estimated and the geometric properties of the trajectory of EM iterations are not well-understood. In this paper, first, using Bessel functions we provide explicit closed-form expressions for the EM updates under all SNR regimes. Then, in the noiseless setting, we completely characterize the behavior of EM iterations by deriving a recurrence relation at the population level and notably show that all the iterations lie on a certain cycloid. Based on this new trajectory-based analysis, we exhibit the theoretical estimate for the exponent of super-linear convergence and further improve the statistical error bound at the finite-sample level. Our analysis provides a new framework for studying the behavior of EM for Mixed Linear Regression., Comment: This paper was accepted by the 41st International Conference on Machine Learning (ICML 2024). The code for numerical experiments is available at https://github.com/dassein/cycloid_em_mlr

There has been considerable effort to better understand the generalization capabilities of deep neural networks both as a means to unlock a theoretical understanding of their success as well as providing directions for further improvements. In this paper, we investigate margin-based multiclass generalization bounds for neural networks which rely on a recent complexity measure, the geometric complexity, developed for neural networks. We derive a new upper bound on the generalization error which scales with the margin-normalized geometric complexity of the network and which holds for a broad family of data distributions and model classes. Our generalization bound is empirically investigated for a ResNet-18 model trained with SGD on the CIFAR-10 and CIFAR-100 datasets with both original and random labels., Comment: Accepted as an ICML 2023 workshop paper (Topology, Algebra and Geometry in Machine Learning)

This paper investigates the conformal isometry hypothesis as a potential explanation for the emergence of hexagonal periodic patterns in the response maps of grid cells. The hypothesis posits that the activities of the population of grid cells form a high-dimensional vector in the neural space, representing the agent's self-position in 2D physical space. As the agent moves in the 2D physical space, the vector rotates in a 2D manifold in the neural space, driven by a recurrent neural network. The conformal isometry hypothesis proposes that this 2D manifold in the neural space is a conformally isometric embedding of the 2D physical space, in the sense that local displacements of the vector in neural space are proportional to local displacements of the agent in the physical space. Thus the 2D manifold forms an internal map of the 2D physical space, equipped with an internal metric. In this paper, we conduct numerical experiments to show that this hypothesis underlies the hexagon periodic patterns of grid cells. We also conduct theoretical analysis to further support this hypothesis. In addition, we propose a conformal modulation of the input velocity of the agent so that the recurrent neural network of grid cells satisfies the conformal isometry hypothesis automatically. To summarize, our work provides numerical and theoretical evidences for the conformal isometry hypothesis for grid cells and may serve as a foundation for further development of normative models of grid cells and beyond., Comment: arXiv admin note: text overlap with arXiv:2310.19192

This paper studies the transfer reinforcement learning (RL) problem where multiple RL problems have different reward functions but share the same underlying transition dynamics. In this setting, the Q-function of each RL problem (task) can be decomposed into a successor feature (SF) and a reward mapping: the former characterizes the transition dynamics, and the latter characterizes the task-specific reward function. This Q-function decomposition, coupled with a policy improvement operator known as generalized policy improvement (GPI), reduces the sample complexity of finding the optimal Q-function, and thus the SF \& GPI framework exhibits promising empirical performance compared to traditional RL methods like Q-learning. However, its theoretical foundations remain largely unestablished, especially when learning the successor features using deep neural networks (SF-DQN). This paper studies the provable knowledge transfer using SFs-DQN in transfer RL problems. We establish the first convergence analysis with provable generalization guarantees for SF-DQN with GPI. The theory reveals that SF-DQN with GPI outperforms conventional RL approaches, such as deep Q-network, in terms of both faster convergence rate and better generalization. Numerical experiments on real and synthetic RL tasks support the superior performance of SF-DQN \& GPI, aligning with our theoretical findings., Comment: arXiv admin note: text overlap with arXiv:2310.16173

This paper presents two models of neural-networks and their training applicable to neural networks of arbitrary width, depth and topology, assuming only finite-energy neural activations; and a novel representor theory for neural networks in terms of a matrix-valued kernel. The first model is exact (un-approximated) and global, casting the neural network as an elements in a reproducing kernel Banach space (RKBS); we use this model to provide tight bounds on Rademacher complexity. The second model is exact and local, casting the change in neural network function resulting from a bounded change in weights and biases (ie. a training step) in reproducing kernel Hilbert space (RKHS) in terms of a local-intrinsic neural kernel (LiNK). This local model provides insight into model adaptation through tight bounds on Rademacher complexity of network adaptation. We also prove that the neural tangent kernel (NTK) is a first-order approximation of the LiNK kernel. Finally, and noting that the LiNK does not provide a representor theory for technical reasons, we present an exact novel representor theory for layer-wise neural network training with unregularized gradient descent in terms of a local-extrinsic neural kernel (LeNK). This representor theory gives insight into the role of higher-order statistics in neural network training and the effect of kernel evolution in neural-network kernel models. Throughout the paper (a) feedforward ReLU networks and (b) residual networks (ResNet) are used as illustrative examples.

With the proliferation of generative AI and the increasing volume of generative data (also called as synthetic data), assessing the fidelity of generative data has become a critical concern. In this paper, we propose a discriminative approach to estimate the total variation (TV) distance between two distributions as an effective measure of generative data fidelity. Our method quantitatively characterizes the relation between the Bayes risk in classifying two distributions and their TV distance. Therefore, the estimation of total variation distance reduces to that of the Bayes risk. In particular, this paper establishes theoretical results regarding the convergence rate of the estimation error of TV distance between two Gaussian distributions. We demonstrate that, with a specific choice of hypothesis class in classification, a fast convergence rate in estimating the TV distance can be achieved. Specifically, the estimation accuracy of the TV distance is proven to inherently depend on the separation of two Gaussian distributions: smaller estimation errors are achieved when the two Gaussian distributions are farther apart. This phenomenon is also validated empirically through extensive simulations. In the end, we apply this discriminative estimation method to rank fidelity of synthetic image data using the MNIST dataset.

This paper investigates score-based diffusion models when the underlying target distribution is concentrated on or near low-dimensional manifolds within the higher-dimensional space in which they formally reside, a common characteristic of natural image distributions. Despite previous efforts to understand the data generation process of diffusion models, existing theoretical support remains highly suboptimal in the presence of low-dimensional structure, which we strengthen in this paper. For the popular Denoising Diffusion Probabilistic Model (DDPM), we find that the dependency of the error incurred within each denoising step on the ambient dimension $d$ is in general unavoidable. We further identify a unique design of coefficients that yields a converges rate at the order of $O(k^{2}/\sqrt{T})$ (up to log factors), where $k$ is the intrinsic dimension of the target distribution and $T$ is the number of steps. This represents the first theoretical demonstration that the DDPM sampler can adapt to unknown low-dimensional structures in the target distribution, highlighting the critical importance of coefficient design. All of this is achieved by a novel set of analysis tools that characterize the algorithmic dynamics in a more deterministic manner.

Differentiable annealed importance sampling (DAIS), proposed by Geffner & Domke (2021) and Zhang et al. (2021), allows optimizing over the initial distribution of AIS. In this paper, we show that, in the limit of many transitions, DAIS minimizes the symmetrized Kullback-Leibler divergence between the initial and target distribution. Thus, DAIS can be seen as a form of variational inference (VI) as its initial distribution is a parametric fit to an intractable target distribution. We empirically evaluate the usefulness of the initial distribution as a variational distribution on synthetic and real-world data, observing that it often provides more accurate uncertainty estimates than VI (optimizing the reverse KL divergence), importance weighted VI, and Markovian score climbing (optimizing the forward KL divergence)., Comment: 22 pages, including 9 pages of main text and 11 pages of appendix, conference paper at ICML 2024, updated terminology