Human intelligence is characterized not only by the capacity to learn complex skills, but the ability to rapidly adapt and acquire new skills within an ever-changing environment. In this work we study how the learning of modular solutions can allow for effective generalization to both unseen and potentially differently distributed data. Our main postulate is that the combination of task segmentation, modular learning and memory-based ensembling can give rise to generalization on an exponentially growing number of unseen tasks. We provide a concrete instantiation of this idea using a combination of: (1) the Forget-Me-Not Process, for task segmentation and memory based ensembling; and (2) Gated Linear Networks, which in contrast to contemporary deep learning techniques use a modular and local learning mechanism. We demonstrate that this system exhibits a number of desirable continual learning properties: robustness to catastrophic forgetting, no negative transfer and increasing levels of positive transfer as more tasks are seen. We show competitive performance against both offline and online methods on standard continual learning benchmarks.

In online learning problems, exploiting low variance plays an important role in obtaining tight performance guarantees yet is challenging because variances are often not known a priori. Recently, considerable progress has been made by Zhang et al. (2021) where they obtain a variance-adaptive regret bound for linear bandits without knowledge of the variances and a horizon-free regret bound for linear mixture Markov decision processes (MDPs). In this paper, we present novel analyses that improve their regret bounds significantly.

Knowledge distillation is an effective approach to learn compact models (students) with the supervision of large and strong models (teachers). As empirically there exists a strong correlation between the performance of teacher and student models, it is commonly believed that a high performing teacher is preferred. Consequently, practitioners tend to use a well trained network or an ensemble of them as the teacher. In this paper, we observe that an intermediate model, i.e., a checkpoint in the middle of the training procedure, often serves as a better teacher compared to the fully converged model, although the former has much lower accuracy. More surprisingly, a weak snapshot ensemble of several intermediate models from a same training trajectory can outperform a strong ensemble of independently trained and fully converged models, when they are used as teachers. We show that this phenomenon can be partially explained by the information bottleneck principle: the feature representations of intermediate models can have higher mutual information regarding the input, and thus contain more ``dark knowledge'' for effective distillation. We further propose an optimal intermediate teacher selection algorithm based on maximizing the total task-related mutual information. Experiments verify its effectiveness and applicability.

PopSign is a smartphone-based bubble-shooter game that helps hearing parentsof deaf infants learn sign language. To help parents practice their ability to sign,PopSign is integrating sign language recognition as part of its gameplay. Fortraining the recognizer, we introduce the PopSign ASL v1.0 dataset that collectsexamples of 250 isolated American Sign Language (ASL) signs using Pixel 4Asmartphone selfie cameras in a variety of environments. It is the largest publiclyavailable, isolated sign dataset by number of examples and is the first dataset tofocus on one-handed, smartphone signs. We collected over 210,000 examplesat 1944x2592 resolution made by 47 consenting Deaf adult signers for whomAmerican Sign Language is their primary language. We manually reviewed 217,866of these examples, of which 175,023 (approximately 700 per sign) were the signintended for the educational game.

We present a novel methodology aimed at optimizing the application of frozen large language models (LLMs) for resource-intensive vision-language (VL) pre-training. The current paradigm uses visual features as prompts to guide language models, with a focus on determining the most relevant visual features for corresponding text. Our approach diverges by concentrating on the language component, specifically identifying the optimal prompts to align with visual features. We introduce the Prompt-Transformer (P-Former), a model that predicts these ideal prompts, which is trained exclusively on linguistic data, bypassing the need for image-text pairings.

We consider supervised learning problems in which set predictions provide explicit uncertainty estimates. Using Choquet integrals (a.k.a. Lov{á}sz extensions), we propose a convex loss function for nondecreasing subset-valued functions obtained as level sets of a real-valued function. This loss function allows optimal trade-offs between conditional probabilistic coverage and the ''size'' of the set, measured by a non-decreasing submodular function. We also propose several extensions that mimic loss functions and criteria for binary classification with asymmetric losses, and show how to naturally obtain sets with optimized conditional coverage. We derive efficient optimization algorithms, either based on stochastic gradient descent or reweighted least-squares formulations, and illustrate our findings with a series of experiments on synthetic datasets for classification and regression tasks, showing improvements over approaches that aim for marginal coverage.

Transformers can implement both generalizable algorithms (e.g., induction heads) and simple positional shortcuts (e.g., memorizing fixed output positions). In this work, we study how the choice of pretraining data distribution steers a shallow transformer toward one behavior or the other. Focusing on a minimal trigger-output prediction task -- copying the token immediately following a special trigger upon its second occurrence -- we present a rigorous analysis of gradient-based training of a single-layer transformer. In both the infinite and finite sample regimes, we prove a transition in the learned mechanism: if input sequences exhibit sufficient diversity, measured by a low ``max-sum'' ratio of trigger-to-trigger distances, the trained model implements an induction head and generalizes to unseen contexts; by contrast, when this ratio is large, the model resorts to a positional shortcut and fails to generalize out-of-distribution (OOD). We also reveal a trade-off between the pretraining context length and OOD generalization, and derive the optimal pretraining distribution that minimizes computational cost per sample. Finally, we validate our theoretical predictions with controlled synthetic experiments, demonstrating that broadening context distributions robustly induces induction heads and enables OOD generalization. Our results shed light on the algorithmic biases of pretrained transformers and offer conceptual guidelines for data-driven control of their learned behaviors.

Proven methods for teaching the readers who struggle most have been known for decades. Why do we often fail to use them? "There's a window of opportunity to intervene," Mark Seidenberg, a cognitive neuroscientist, said. "You don't want to let that go." In 2024, my niece Caroline received a Ph.D. in gravitational-wave physics. Her research interests include "the impact of model inaccuracies on biases in parameters recovered from gravitational wave data" and "Petrov type, principal null directions, and Killing tensors of slowly rotating black holes in quadratic gravity." I watched a little of her dissertation defense, on Zoom, and was lost as soon as she'd finished introducing herself. She and her husband now live in Italy, where she has a postdoctoral appointment. Caroline's academic achievements seem especially impressive if you know that until third grade she could barely read: to her, words on a page looked like a pulsing mass. She attended a private school in Connecticut, and there was a set time every day when students selected books to read on their own. "I can't remember how long that lasted, but it felt endless," she told me. She hid her disability by turning pages when her classmates did, and by volunteering to draw illustrations during group story-writing projects. One day, she told her grandmother that she could sound out individual letters but when she got to "the end of a row" she couldn't remember what had come before. A psychologist eventually identified her condition as dyslexia. Fluent readers sometimes think of dyslexia as a tendency to put letters in the wrong order or facing the wrong direction, but it's more complicated than that.

A new paradigm in machine learning for scientific computing is focused on designing learning algorithms and methods for continuum problems. This paradigm is referred to as operator learning and has received considerable interest in the last few years [5,7,18,20,23-25,27,30,34,36]. The basic task may be posed as learning a map between infinite-dimensional function spaces, i.e., learning an operator F: X Y, where, for example, X and Y are real, separable Hilbert spaces. Operator learning naturally arises in many scientific problems where one wants to learn how a continuum model, often described by partial differential equations (PDEs), maps inputs, such as parameters or boundary conditions, to outputs, such as states or observables. A prototypical example to keep in mind is learning parameter-to-solution maps of parametric PDEs [1,2,11]. In contrast to more classical surrogate modeling, which typically focuses on learning finite-dimensional parameter-to-solution maps for some fixed discretization, operator learning directly aims to learn/approximate the continuum map F: X Y itself. Thus, the inputs and outputs are functions (not vectors) and the goal is to directly design discretization-invariant methods [7,23]. From a statistical perspective, this naturally leads to a nonparametric regression problem in which both the object of interest (the operator) and the observations (finite number of noisy samples) are infinite-dimensional.

We present a novel theoretical analysis of Federated SARSA (FedSARSA) with linear function approximation and local training. We establish convergence guarantees for FedSARSA in the presence of heterogeneity, both in local transitions and rewards, providing the first sample and communication complexity bounds in this setting. At the core of our analysis is a new, exact multi-step error expansion for single-agent SARSA, which is of independent interest. Our analysis precisely quantifies the impact of heterogeneity, demonstrating the convergence of FedSARSA with multiple local updates. Crucially, we show that FedSARSA achieves linear speed-up with respect to the number of agents, up to higher-order terms due to Markovian sampling. Numerical experiments support our theoretical findings.