Machine Learning (cs.LG)

Abstract: In recent years, research on learning with noisy labels has focused on devising novel algorithms that can achieve robustness to noisy training labels while generalizing to clean data. These algorithms often incorporate sophisticated techniques, such as noise modeling, label correction, and co-training. In this study, we demonstrate that a simple baseline using cross-entropy loss, combined with widely used regularization strategies like learning rate decay, model weights average, and data augmentations, can outperform state-of-the-art methods. Our findings suggest that employing a combination of regularization strategies can be more effective than intricate algorithms in tackling the challenges of learning with noisy labels. While some of these regularization strategies have been utilized in previous noisy label learning research, their full potential has not been thoroughly explored. Our results encourage a reevaluation of benchmarks for learning with noisy labels and prompt reconsideration of the role of specialized learning algorithms designed for training with noisy labels.

Abstract: This thesis explores open-sourced machine learning (ML) model explanation tools to understand whether these tools can allow a layman to visualize, understand, and suggest intuitive remedies to unfairness in ML-based decision-support systems. Machine learning models trained on datasets biased against minority groups are increasingly used to guide life-altering social decisions, prompting the urgent need to study their logic for unfairness. Due to this problem's impact on vast populations of the general public, it is critical for the layperson -- not just subject matter experts in social justice or machine learning experts -- to understand the nature of unfairness within these algorithms and the potential trade-offs. Existing research on fairness in machine learning focuses mostly on the mathematical definitions and tools to understand and remedy unfair models, with some directly citing user-interactive tools as necessary for future work. This thesis presents FairLay-ML, a proof-of-concept GUI integrating some of the most promising tools to provide intuitive explanations for unfair logic in ML models by integrating existing research tools (e.g. Local Interpretable Model-Agnostic Explanations) with existing ML-focused GUI (e.g. Python Streamlit). We test FairLay-ML using models of various accuracy and fairness generated by an unfairness detector tool, Parfait-ML, and validate our results using Themis. Our study finds that the technology stack used for FairLay-ML makes it easy to install and provides real-time black-box explanations of pre-trained models to users. Furthermore, the explanations provided translate to actionable remedies.

Abstract: The labor-intensive annotation process of semantic segmentation datasets is often prone to errors, since humans struggle to label every pixel correctly. We study algorithms to automatically detect such annotation errors, in particular methods to score label quality, such that the images with the lowest scores are least likely to be correctly labeled. This helps prioritize what data to review in order to ensure a high-quality training/evaluation dataset, which is critical in sensitive applications such as medical imaging and autonomous vehicles. Widely applicable, our label quality scores rely on probabilistic predictions from a trained segmentation model -- any model architecture and training procedure can be utilized. Here we study 7 different label quality scoring methods used in conjunction with a DeepLabV3+ or a FPN segmentation model to detect annotation errors in a version of the SYNTHIA dataset. Precision-recall evaluations reveal a score -- the soft-minimum of the model-estimated likelihoods of each pixel's annotated class -- that is particularly effective to identify images that are mislabeled, across multiple types of annotation error.

Abstract: Explainable Artificial Intelligence (XAI) has gained significant attention recently as the demand for transparency and interpretability of machine learning models has increased. In particular, XAI for time series data has become increasingly important in finance, healthcare, and climate science. However, evaluating the quality of explanations, such as attributions provided by XAI techniques, remains challenging. This paper provides an in-depth analysis of using perturbations to evaluate attributions extracted from time series models. A perturbation analysis involves systematically modifying the input data and evaluating the impact on the attributions generated by the XAI method. We apply this approach to several state-of-the-art XAI techniques and evaluate their performance on three time series classification datasets. Our results demonstrate that the perturbation analysis approach can effectively evaluate the quality of attributions and provide insights into the strengths and limitations of XAI techniques. Such an approach can guide the selection of XAI methods for time series data, e.g., focusing on return time rather than precision, and facilitate the development of more reliable and interpretable machine learning models for time series analysis.

Abstract: Given a sequence of observable variables $\{(x_1, y_1), \ldots, (x_n, y_n)\}$, the conformal prediction method estimates a confidence set for $y_{n+1}$ given $x_{n+1}$ that is valid for any finite sample size by merely assuming that the joint distribution of the data is permutation invariant. Although attractive, computing such a set is computationally infeasible in most regression problems. Indeed, in these cases, the unknown variable $y_{n+1}$ can take an infinite number of possible candidate values, and generating conformal sets requires retraining a predictive model for each candidate. In this paper, we focus on a sparse linear model with only a subset of variables for prediction and use numerical continuation techniques to approximate the solution path efficiently. The critical property we exploit is that the set of selected variables is invariant under a small perturbation of the input data. Therefore, it is sufficient to enumerate and refit the model only at the change points of the set of active features and smoothly interpolate the rest of the solution via a Predictor-Corrector mechanism. We show how our path-following algorithm accurately approximates conformal prediction sets and illustrate its performance using synthetic and real data examples.

Abstract: How to obtain informative representations of transactions and then perform the identification of fraudulent transactions is a crucial part of ensuring financial security. Recent studies apply Graph Neural Networks (GNNs) to the transaction fraud detection problem. Nevertheless, they encounter challenges in effectively learning spatial-temporal information due to structural limitations. Moreover, few prior GNN-based detectors have recognized the significance of incorporating global information, which encompasses similar behavioral patterns and offers valuable insights for discriminative representation learning. Therefore, we propose a novel heterogeneous graph neural network called Spatial-Temporal-Aware Graph Transformer (STA-GT) for transaction fraud detection problems. Specifically, we design a temporal encoding strategy to capture temporal dependencies and incorporate it into the graph neural network framework, enhancing spatial-temporal information modeling and improving expressive ability. Furthermore, we introduce a transformer module to learn local and global information. Pairwise node-node interactions overcome the limitation of the GNN structure and build up the interactions with the target node and long-distance ones. Experimental results on two financial datasets compared to general GNN models and GNN-based fraud detectors demonstrate that our proposed method STA-GT is effective on the transaction fraud detection task.

Abstract: Due to their dynamic properties such as irregular sampling rate and high-frequency sampling, Continuous Time Series (CTS) are found in many applications. Since CTS with irregular sampling rate are difficult to model with standard Recurrent Neural Networks (RNNs), RNNs have been generalised to have continuous-time hidden dynamics defined by a Neural Ordinary Differential Equation (Neural ODE), leading to the ODE-RNN model. Another approach that provides a better modelling is that of the Latent ODE model, which constructs a continuous-time model where a latent state is defined at all times. The Latent ODE model uses a standard RNN as the encoder and a Neural ODE as the decoder. However, since the RNN encoder leads to difficulties with missing data and ill-defined latent variables, a Latent ODE-RNN model has recently been proposed that uses a ODE-RNN model as the encoder instead. Both the Latent ODE and Latent ODE-RNN models are difficult to train due to the vanishing and exploding gradients problem. To overcome this problem, the main contribution of this paper is to propose and illustrate a new model based on a new Latent ODE using an ODE-LSTM (Long Short-Term Memory) network as an encoder -- the Latent ODE-LSTM model. To limit the growth of the gradients the Norm Gradient Clipping strategy was embedded on the Latent ODE-LSTM model. The performance evaluation of the new Latent ODE-LSTM (with and without Norm Gradient Clipping) for modelling CTS with regular and irregular sampling rates is then demonstrated. Numerical experiments show that the new Latent ODE-LSTM performs better than Latent ODE-RNNs and can avoid the vanishing and exploding gradients during training.

Abstract: Energy storage solutions play an increasingly important role in modern infrastructure and lead-acid batteries are among the most commonly used in the rechargeable category. Due to normal degradation over time, correctly determining the battery's State of Health (SoH) and Remaining Useful Life (RUL) contributes to enhancing predictive maintenance, reliability, and longevity of battery systems. Besides improving the cost savings, correct estimation of the SoH can lead to reduced pollution though reuse of retired batteries. This paper presents a mapping study of the state-of-the-art in machine learning methods for estimating the SoH and RUL of lead-acid batteries. These two indicators are critical in the battery management systems of electric vehicles, renewable energy systems, and other applications that rely heavily on this battery technology. In this study, we analyzed the types of machine learning algorithms employed for estimating SoH and RUL, and evaluated their performance in terms of accuracy and inference time. Additionally, this mapping identifies and analyzes the most commonly used combinations of sensors in specific applications, such as vehicular batteries. The mapping concludes by highlighting potential gaps and opportunities for future research, which lays the foundation for further advancements in the field.

Abstract: We present a simple linear regression based approach for learning the weights and biases of a neural network, as an alternative to standard gradient based backpropagation. The present work is exploratory in nature, and we restrict the description and experiments to (i) simple feedforward neural networks, (ii) scalar (single output) regression problems, and (iii) invertible activation functions. However, the approach is intended to be extensible to larger, more complex architectures. The key idea is the observation that the input to every neuron in a neural network is a linear combination of the activations of neurons in the previous layer, as well as the parameters (weights and biases) of the layer. If we are able to compute the ideal total input values to every neuron by working backwards from the output, we can formulate the learning problem as a linear least squares problem which iterates between updating the parameters and the activation values. We present an explicit algorithm that implements this idea, and we show that (at least for simple problems) the approach is more stable and faster than gradient-based backpropagation.

Abstract: Several membership inference (MI) attacks have been proposed to audit a target DNN. Given a set of subjects, MI attacks tell which subjects the target DNN has seen during training. This work focuses on the post-training MI attacks emphasizing high confidence membership detection -- True Positive Rates (TPR) at low False Positive Rates (FPR). Current works in this category -- likelihood ratio attack (LiRA) and enhanced MI attack (EMIA) -- only perform well on complex datasets (e.g., CIFAR-10 and Imagenet) where the target DNN overfits its train set, but perform poorly on simpler datasets (0% TPR by both attacks on Fashion-MNIST, 2% and 0% TPR respectively by LiRA and EMIA on MNIST at 1% FPR). To address this, firstly, we unify current MI attacks by presenting a framework divided into three stages -- preparation, indication and decision. Secondly, we utilize the framework to propose two novel attacks: (1) Adversarial Membership Inference Attack (AMIA) efficiently utilizes the membership and the non-membership information of the subjects while adversarially minimizing a novel loss function, achieving 6% TPR on both Fashion-MNIST and MNIST datasets; and (2) Enhanced AMIA (E-AMIA) combines EMIA and AMIA to achieve 8% and 4% TPRs on Fashion-MNIST and MNIST datasets respectively, at 1% FPR. Thirdly, we introduce two novel augmented indicators that positively leverage the loss information in the Gaussian neighborhood of a subject. This improves TPR of all four attacks on average by 2.5% and 0.25% respectively on Fashion-MNIST and MNIST datasets at 1% FPR. Finally, we propose simple, yet novel, evaluation metric, the running TPR average (RTA) at a given FPR, that better distinguishes different MI attacks in the low FPR region. We also show that AMIA and E-AMIA are more transferable to the unknown DNNs (other than the target DNN) and are more robust to DP-SGD training as compared to LiRA and EMIA.

Abstract: The optimal prediction strategy for out-of-distribution (OOD) setups is a fundamental question in machine learning. In this paper, we address this question and present several contributions. We propose three reject option models for OOD setups: the Cost-based model, the Bounded TPR-FPR model, and the Bounded Precision-Recall model. These models extend the standard reject option models used in non-OOD setups and define the notion of an optimal OOD selective classifier. We establish that all the proposed models, despite their different formulations, share a common class of optimal strategies. Motivated by the optimal strategy, we introduce double-score OOD methods that leverage uncertainty scores from two chosen OOD detectors: one focused on OOD/ID discrimination and the other on misclassification detection. The experimental results consistently demonstrate the superior performance of this simple strategy compared to state-of-the-art methods. Additionally, we propose novel evaluation metrics derived from the definition of the optimal strategy under the proposed OOD rejection models. These new metrics provide a comprehensive and reliable assessment of OOD methods without the deficiencies observed in existing evaluation approaches.

Abstract: Many real-world optimization problems contain unknown parameters that must be predicted prior to solving. To train the predictive machine learning (ML) models involved, the commonly adopted approach focuses on maximizing predictive accuracy. However, this approach does not always lead to the minimization of the downstream task loss. Decision-focused learning (DFL) is a recently proposed paradigm whose goal is to train the ML model by directly minimizing the task loss. However, state-of-the-art DFL methods are limited by the assumptions they make about the structure of the optimization problem (e.g., that the problem is linear) and by the fact that can only predict parameters that appear in the objective function. In this work, we address these limitations by instead predicting \textit{distributions} over parameters and adopting score function gradient estimation (SFGE) to compute decision-focused updates to the predictive model, thereby widening the applicability of DFL. Our experiments show that by using SFGE we can: (1) deal with predictions that occur both in the objective function and in the constraints; and (2) effectively tackle two-stage stochastic optimization problems.

Abstract: Graph neural networks have become the standard approach for dealing with learning problems on graphs. Among the different variants of graph neural networks, graph attention networks (GATs) have been applied with great success to different tasks. In the GAT model, each node assigns an importance score to its neighbors using an attention mechanism. However, similar to other graph neural networks, GATs aggregate messages from nodes that belong to different classes, and therefore produce node representations that are not well separated with respect to the different classes, which might hurt their performance. In this work, to alleviate this problem, we propose a new technique that can be incorporated into any graph attention model to encourage higher attention scores between nodes that share the same class label. We evaluate the proposed method on several node classification datasets demonstrating increased performance over standard baseline models.

Abstract: Artificial intelligence has achieved significant success in handling complex tasks in recent years. This success is due to advances in machine learning algorithms and hardware acceleration. In order to obtain more accurate results and solve more complex problems, algorithms must be trained with more data. This huge amount of data could be time-consuming to process and require a great deal of computation. This solution could be achieved by distributing the data and algorithm across several machines, which is known as distributed machine learning. There has been considerable effort put into distributed machine learning algorithms, and different methods have been proposed so far. In this article, we present a comprehensive summary of the current state-of-the-art in the field through the review of these algorithms. We divide this algorithms in classification and clustering (traditional machine learning), deep learning and deep reinforcement learning groups. Distributed deep learning has gained more attention in recent years and most of studies worked on this algorithms. As a result, most of the articles we discussed here belong to this category. Based on our investigation of algorithms, we highlight limitations that should be addressed in future research.

Abstract: Imagine a learner L who tries to infer a hidden concept from a collection of observations. Building on the work [4] of Ferri et al., we assume the learner to be parameterized by priors P(c) and by c-conditional likelihoods P(z|c) where c ranges over all concepts in a given class C and z ranges over all observations in an observation set Z. L is called a MAP-learner (resp. an MLE-learner) if it thinks of a collection S of observations as a random sample and returns the concept with the maximum a-posteriori probability (resp. the concept which maximizes the c-conditional likelihood of S). Depending on whether L assumes that S is obtained from ordered or unordered sampling resp. from sampling with or without replacement, we can distinguish four different sampling modes. Given a target concept c in C, a teacher for a MAP-learner L aims at finding a smallest collection of observations that causes L to return c. This approach leads in a natural manner to various notions of a MAP- or MLE-teaching dimension of a concept class C. Our main results are: We show that this teaching model has some desirable monotonicity properties. We clarify how the four sampling modes are related to each other. As for the (important!) special case, where concepts are subsets of a domain and observations are 0,1-labeled examples, we obtain some additional results. First of all, we characterize the MAP- and MLE-teaching dimension associated with an optimally parameterized MAP-learner graph-theoretically. From this central result, some other ones are easy to derive. It is shown, for instance, that the MLE-teaching dimension is either equal to the MAP-teaching dimension or exceeds the latter by 1. It is shown furthermore that these dimensions can be bounded from above by the so-called antichain number, the VC-dimension and related combinatorial parameters. Moreover they can be computed in polynomial time.

Abstract: The aging population has led to a growing number of falls in our society, affecting global public health worldwide. This paper presents CareFall, an automatic Fall Detection System (FDS) based on wearable devices and Artificial Intelligence (AI) methods. CareFall considers the accelerometer and gyroscope time signals extracted from a smartwatch. Two different approaches are used for feature extraction and classification: i) threshold-based, and ii) machine learning-based. Experimental results on two public databases show that the machine learning-based approach, which combines accelerometer and gyroscope information, outperforms the threshold-based approach in terms of accuracy, sensitivity, and specificity. This research contributes to the design of smart and user-friendly solutions to mitigate the negative consequences of falls among older people.

Abstract: Different distribution shifts require different algorithmic and operational interventions. Methodological research must be grounded by the specific shifts they address. Although nascent benchmarks provide a promising empirical foundation, they implicitly focus on covariate shifts, and the validity of empirical findings depends on the type of shift, e.g., previous observations on algorithmic performance can fail to be valid when the $Y|X$ distribution changes. We conduct a thorough investigation of natural shifts in 5 tabular datasets over 86,000 model configurations, and find that $Y|X$-shifts are most prevalent. To encourage researchers to develop a refined language for distribution shifts, we build WhyShift, an empirical testbed of curated real-world shifts where we characterize the type of shift we benchmark performance over. Since $Y|X$-shifts are prevalent in tabular settings, we identify covariate regions that suffer the biggest $Y|X$-shifts and discuss implications for algorithmic and data-based interventions. Our testbed highlights the importance of future research that builds an understanding of how distributions differ.

Abstract: The time evolution of physical systems is described by differential equations, which depend on abstract quantities like energy and force. Traditionally, these quantities are derived as functionals based on observables such as positions and velocities. Discovering these governing symbolic laws is the key to comprehending the interactions in nature. Here, we present a Hamiltonian graph neural network (HGNN), a physics-enforced GNN that learns the dynamics of systems directly from their trajectory. We demonstrate the performance of HGNN on n-springs, n-pendulums, gravitational systems, and binary Lennard Jones systems; HGNN learns the dynamics in excellent agreement with the ground truth from small amounts of data. We also evaluate the ability of HGNN to generalize to larger system sizes, and to hybrid spring-pendulum system that is a combination of two original systems (spring and pendulum) on which the models are trained independently. Finally, employing symbolic regression on the learned HGNN, we infer the underlying equations relating the energy functionals, even for complex systems such as the binary Lennard-Jones liquid. Our framework facilitates the interpretable discovery of interaction laws directly from physical system trajectories. Furthermore, this approach can be extended to other systems with topology-dependent dynamics, such as cells, polydisperse gels, or deformable bodies.

Abstract: Federated learning has become a popular method to learn from decentralized heterogeneous data. Federated semi-supervised learning (FSSL) emerges to train models from a small fraction of labeled data due to label scarcity on decentralized clients. Existing FSSL methods assume independent and identically distributed (IID) labeled data across clients and consistent class distribution between labeled and unlabeled data within a client. This work studies a more practical and challenging scenario of FSSL, where data distribution is different not only across clients but also within a client between labeled and unlabeled data. To address this challenge, we propose a novel FSSL framework with dual regulators, FedDure.} FedDure lifts the previous assumption with a coarse-grained regulator (C-reg) and a fine-grained regulator (F-reg): C-reg regularizes the updating of the local model by tracking the learning effect on labeled data distribution; F-reg learns an adaptive weighting scheme tailored for unlabeled instances in each client. We further formulate the client model training as bi-level optimization that adaptively optimizes the model in the client with two regulators. Theoretically, we show the convergence guarantee of the dual regulators. Empirically, we demonstrate that FedDure is superior to the existing methods across a wide range of settings, notably by more than 11% on CIFAR-10 and CINIC-10 datasets.

Abstract: Production deployments in complex systems require ML architectures to be highly efficient and usable against multiple tasks. Particularly demanding are classification problems in which data arrives in a streaming fashion and each class is presented separately. Recent methods with stochastic gradient learning have been shown to struggle in such setups or have limitations like memory buffers, and being restricted to specific domains that disable its usage in real-world scenarios. For this reason, we present a fully differentiable architecture based on the Mixture of Experts model, that enables the training of high-performance classifiers when examples from each class are presented separately. We conducted exhaustive experiments that proved its applicability in various domains and ability to learn online in production environments. The proposed technique achieves SOTA results without a memory buffer and clearly outperforms the reference methods.

Abstract: Machine learning for differential equations paves the way for computationally efficient alternatives to numerical solvers, with potentially broad impacts in science and engineering. Though current algorithms typically require simulated training data tailored to a given setting, one may instead wish to learn useful information from heterogeneous sources, or from real dynamical systems observations that are messy or incomplete. In this work, we learn general-purpose representations of PDEs from heterogeneous data by implementing joint embedding methods for self-supervised learning (SSL), a framework for unsupervised representation learning that has had notable success in computer vision. Our representation outperforms baseline approaches to invariant tasks, such as regressing the coefficients of a PDE, while also improving the time-stepping performance of neural solvers. We hope that our proposed methodology will prove useful in the eventual development of general-purpose foundation models for PDEs.

Abstract: Multimodal learning models have become increasingly important as they surpass single-modality approaches on diverse tasks ranging from question-answering to autonomous driving. Despite the importance of multimodal learning, existing efforts focus on NLP applications, where the number of modalities is typically less than four (audio, video, text, images). However, data inputs in other domains, such as the medical field, may include X-rays, PET scans, MRIs, genetic screening, clinical notes, and more, creating a need for both efficient and accurate information fusion. Many state-of-the-art models rely on pairwise cross-modal attention, which does not scale well for applications with more than three modalities. For $n$ modalities, computing attention will result in $n \choose 2$ operations, potentially requiring considerable amounts of computational resources. To address this, we propose a new domain-neutral attention mechanism, One-Versus-Others (OvO) attention, that scales linearly with the number of modalities and requires only $n$ attention operations, thus offering a significant reduction in computational complexity compared to existing cross-modal attention algorithms. Using three diverse real-world datasets as well as an additional simulation experiment, we show that our method improves performance compared to popular fusion techniques while decreasing computation costs.

Abstract: Denoising diffusion models have recently emerged as the predominant paradigm for generative modelling. Their extension to Riemannian manifolds has facilitated their application to an array of problems in the natural sciences. Yet, in many practical settings, such manifolds are defined by a set of constraints and are not covered by the existing (Riemannian) diffusion model methodology. Recent work has attempted to address this issue by employing novel noising processes based on logarithmic barrier methods or reflected Brownian motions. However, the associated samplers are computationally burdensome as the complexity of the constraints increases. In this paper, we introduce an alternative simple noising scheme based on Metropolis sampling that affords substantial gains in computational efficiency and empirical performance compared to the earlier samplers. Of independent interest, we prove that this new process corresponds to a valid discretisation of the reflected Brownian motion. We demonstrate the scalability and flexibility of our approach on a range of problem settings with convex and non-convex constraints, including applications from geospatial modelling, robotics and protein design.