Machine learning (ML) provides powerful tools for predictive modeling. ML's popularity stems from the promise of sample-level prediction with applications across a variety of fields from physics and marketing to healthcare. However, if not properly implemented and evaluated, ML pipelines may contain leakage typically resulting in overoptimistic performance estimates and failure to generalize to new data. This can have severe negative financial and societal implications. Our aim is to expand understanding associated with causes leading to leakage when designing, implementing, and evaluating ML pipelines. Illustrated by concrete examples, we provide a comprehensive overview and discussion of various types of leakage that may arise in ML pipelines.

It is often used to justify the effectiveness of machine learning algorithms in high-dimensional settings, since the curse of dimensionality can be circumvented if the data concentrates on a lowdimensional manifold. It is, however, evident that several low-dimensional (and hence, visualisable) datasets do not satisfy the Manifold Hypothesis. Instead, such data can have singularities -- points at which the local geometry does not resemble n-dimensional Euclidean space for any n. Prime examples of singular loci of datasets include branching points in neurons and cosmic filaments. Furthermore, standard image datasets (such as MNIST and CIFAR-10) are known to have non-constant intrinsic dimension [17], whereas a connected manifold must possess the same intrinsic dimension throughout. Whenever such non-manifold behaviour within datasets is of interest, it becomes natural to wonder whether it can be accurately and automatically identified. Particularly in large, high-dimensional datasets where visual inspection is impossible, we seek tools to identify and locate singularities within datasets. Our focus here is on unsupervised singularity detection, where one has recourse neither to a plethora of training data, nor the opportunity to regenerate samples along an unknown probability measure.

Visual object navigation using learning methods is one of the key tasks in mobile robotics. This paper introduces a new representation of a scene semantic map formed during the embodied agent interaction with the indoor environment. It is based on a neural network method that adjusts the weights of the segmentation model with backpropagation of the predicted fusion loss values during inference on a regular (backward) or delayed (forward) image sequence. We have implemented this representation into a full-fledged navigation approach called SkillTron, which can select robot skills from end-to-end policies based on reinforcement learning and classic map-based planning methods. The proposed approach makes it possible to form both intermediate goals for robot exploration and the final goal for object navigation. We conducted intensive experiments with the proposed approach in the Habitat environment, which showed a significant superiority in navigation quality metrics compared to state-of-the-art approaches.

The rapid development of language models (LMs) brings unprecedented accessibility and usage for both models and users. On the one hand, powerful LMs, trained with massive textual data, achieve state-of-the-art performance over numerous downstream NLP tasks. On the other hand, more and more attention is paid to unrestricted model accesses that may bring malicious privacy risks of data leakage. To address these issues, many recent works propose privacy-preserving language models (PPLMs) with differential privacy (DP). Unfortunately, different DP implementations make it challenging for a fair comparison among existing PPLMs. In this paper, we present P-Bench, a multi-perspective privacy evaluation benchmark to empirically and intuitively quantify the privacy leakage of LMs. Instead of only protecting and measuring the privacy of protected data with DP parameters, P-Bench sheds light on the neglected inference data privacy during actual usage. P-Bench first clearly defines multi-faceted privacy objectives during private fine-tuning. Then, P-Bench constructs a unified pipeline to perform private fine-tuning. Lastly, P-Bench performs existing privacy attacks on LMs with pre-defined privacy objectives as the empirical evaluation results. The empirical attack results are used to fairly and intuitively evaluate the privacy leakage of various PPLMs. We conduct extensive experiments on three datasets of GLUE for mainstream LMs.

Language models have been shown to perform remarkably well on a wide range of natural language processing tasks. In this paper, we propose LEAP, a novel system that uses language models to perform multi-step logical reasoning and incorporates explicit planning into the inference procedure. Explicit planning enables the system to make more informed reasoning decisions at each step by looking ahead into their future effects. Moreover, we propose a training strategy that safeguards the planning process from being led astray by spurious features. Our full system significantly outperforms other competing methods on multiple standard datasets. When using small T5 models as its core selection and deduction components, our system performs competitively compared to GPT-3 despite having only about 1B parameters (i.e., 175 times smaller than GPT-3). When using GPT-3.5, it significantly outperforms chain-of-thought prompting on the challenging PrOntoQA dataset. We have conducted extensive empirical studies to demonstrate that explicit planning plays a crucial role in the system's performance.

Background and Objectives: Reproducibility is a major challenge in developing machine learning (ML)-based solutions in computational pathology (CompPath). The NCI Imaging Data Commons (IDC) provides >120 cancer image collections according to the FAIR principles and is designed to be used with cloud ML services. Here, we explore its potential to facilitate reproducibility in CompPath research. Methods: Using the IDC, we implemented two experiments in which a representative ML-based method for classifying lung tumor tissue was trained and/or evaluated on different datasets. To assess reproducibility, the experiments were run multiple times with separate but identically configured instances of common ML services. Results: The AUC values of different runs of the same experiment were generally consistent. However, we observed small variations in AUC values of up to 0.045, indicating a practical limit to reproducibility. Conclusions: We conclude that the IDC facilitates approaching the reproducibility limit of CompPath research (i) by enabling researchers to reuse exactly the same datasets and (ii) by integrating with cloud ML services so that experiments can be run in identically configured computing environments.

Maximum mean discrepancy (MMD) refers to a general class of nonparametric two-sample tests that are based on maximizing the mean difference over samples from one distribution $P$ versus another $Q$, over all choices of data transformations $f$ living in some function space $\mathcal{F}$. Inspired by recent work that connects what are known as functions of $\textit{Radon bounded variation}$ (RBV) and neural networks (Parhi and Nowak, 2021, 2023), we study the MMD defined by taking $\mathcal{F}$ to be the unit ball in the RBV space of a given smoothness order $k \geq 0$. This test, which we refer to as the $\textit{Radon-Kolmogorov-Smirnov}$ (RKS) test, can be viewed as a generalization of the well-known and classical Kolmogorov-Smirnov (KS) test to multiple dimensions and higher orders of smoothness. It is also intimately connected to neural networks: we prove that the witness in the RKS test -- the function $f$ achieving the maximum mean difference -- is always a ridge spline of degree $k$, i.e., a single neuron in a neural network. This allows us to leverage the power of modern deep learning toolkits to (approximately) optimize the criterion that underlies the RKS test. We prove that the RKS test has asymptotically full power at distinguishing any distinct pair $P \not= Q$ of distributions, derive its asymptotic null distribution, and carry out extensive experiments to elucidate the strengths and weakenesses of the RKS test versus the more traditional kernel MMD test.

Out-of-distribution (OOD) detection is essential for the safe deployment of AI. Particularly, OOD detectors should generalize effectively across diverse scenarios. To improve upon the generalizability of existing OOD detectors, we introduce a highly versatile OOD detector, called Neural Collapse inspired OOD detector (NC-OOD). We extend the prevalent observation that in-distribution (ID) features tend to form clusters, whereas OOD features are far away. Particularly, based on the recent observation, Neural Collapse, we further demonstrate that ID features tend to cluster in proximity to weight vectors. From our extended observation, we propose to detect OOD based on feature proximity to weight vectors. To further rule out OOD samples, we leverage the observation that OOD features tend to reside closer to the origin than ID features. Extensive experiments show that our approach enhances the generalizability of existing work and can consistently achieve state-of-the-art OOD detection performance across a wide range of OOD Benchmarks over different classification tasks, training losses, and model architectures. Machine learning models deployed in practice will inevitably encounter samples that deviate from the training distribution. As a classifier cannot make meaningful predictions on test samples that belong to unseen classes during training, it is important to actively detect and handle Out-of-Distribution (OOD) samples. Considering the diverse application scenarios, an effective OOD detector should generalize across classification tasks of different input resolutions, number of classes, classification accuracy, as well as classifiers under different training schemes and architectures. Since Nguyen et al. (2015) reveals that neural networks tend to be over-confident on OOD samples, an extensive body of research has been focused on developing effective OOD detection algorithms.

Biometry has proved its capability in terms of recognition accuracy. Now, it is widely used for automated border control with the biometric passport, to unlock a smartphone or a computer with a fingerprint or a face recognition algorithm. While identity verification is widely democratized, pure identification with no additional clues is still a work in progress. The identification difficulty depends on the population size, as the larger the group is, the larger the confusion risk. For collision prevention, biometric traits must be sufficiently distinguishable to scale to considerable groups, and algorithms should be able to capture their differences accurately. Most biometric works are purely experimental, and it is impossible to extrapolate the results to a smaller or a larger group. In this work, we propose a theoretical analysis of the distinguishability problem, which governs the error rates of biometric systems. We demonstrate simple relationships between the population size and the number of independent bits necessary to prevent collision in the presence of noise. This work provides the lowest lower bound for memory requirements. The results are very encouraging, as the biometry of the whole Earth population can fit in a regular disk, leaving some space for noise and redundancy.

Supervised learning models have been used in various domains such as lending, college admission, face recognition, natural language processing, etc. However, they may inherit pre-existing biases from training data and exhibit discrimination against protected social groups. Various fairness notions have been proposed to address unfairness issues. In this work, we focus on Equalized Loss (EL), a fairness notion that requires the expected loss to be (approximately) equalized across different groups. Imposing EL on the learning process leads to a non-convex optimization problem even if the loss function is convex, and the existing fair learning algorithms cannot properly be adopted to find the fair predictor under the EL constraint. This paper introduces an algorithm that can leverage off-the-shelf convex programming tools (e.g., CVXPY) to efficiently find the global optimum of this non-convex optimization. In particular, we propose the ELminimizer algorithm, which finds the optimal fair predictor under EL by reducing the non-convex optimization to a sequence of convex optimization problems. We theoretically prove that our algorithm finds the global optimal solution under certain conditions. Then, we support our theoretical results through several empirical studies.