Machine Learning

FKEA: a novel, scalable method for reference-free evaluation of generative models’ diversity using random Fourier features, overcoming computational limitations of existing entropy-based scores.

A new self-supervised learning approach, Multi-task Masking (MtM), significantly improves the prediction accuracy of neural population activity by capturing neural dynamics at multiple spatial scales,…

Gradient EM for over-parameterized Gaussian Mixture Models globally converges with a sublinear rate, solving a longstanding open problem in machine learning.

A new dual-learning model resolves the paradox of perceptual learning, showing how task-based and feature-based learning interact to produce both specific and transferable improvements in sensory perc…

TinyTTA enables efficient test-time adaptation on memory-constrained edge devices using a novel self-ensemble and early-exit strategy, improving accuracy and reducing memory usage.

Tiny Time Mixers (TTMs) achieve state-of-the-art zero/few-shot multivariate time series forecasting, outperforming existing benchmarks while drastically reducing computational requirements.

TimeXer empowers transformers for superior time series forecasting by cleverly integrating exogenous variables, achieving state-of-the-art results on diverse benchmarks.

TIME-FFM: a Federated Foundation Model empowers time series forecasting using pre-trained Language Models, tackling data scarcity and privacy concerns for superior few-shot and zero-shot predictions.

Time-Constrained Robust MDPs (TC-RMDPs) improve reinforcement learning by addressing limitations of traditional methods, offering a novel framework for handling real-world uncertainties and yielding m…

Networks trained on continuous sensory data spontaneously develop place cell-like responses, demonstrating that time-encoded experience can create spatial maps in the brain.

W-DBO efficiently tackles stale data in dynamic Bayesian Optimization by leveraging a novel Wasserstein distance-based criterion to remove irrelevant observations, maintaining high sampling frequency …

This research enhances meta-learning robustness by theoretically grounding and practically improving tail-risk minimization, achieving improved fast adaptation in the task space.

PCA tree: a novel hierarchical dimensionality reduction model visualized using oblique trees and local PCAs, offering speed and interpretability.

Contrary to expectations, pre-trained visual representations surprisingly don’t improve model-based reinforcement learning’s sample efficiency or generalization; data diversity and network architectu…

TD learning surprisingly outperforms Monte Carlo methods for rare event prediction in Markov chains, achieving relative accuracy with polynomially, instead of exponentially, many observed transitions.

Star geometry reveals optimal data-driven regularizers!

This paper introduces a selective G-Bispectrum algorithm, slashing the computational complexity from O(|G|^2) to O(|G|), making G-invariant deep learning faster and more scalable.

Neural networks exhibit ‘Neural Regression Collapse’ (NRC) during training, where feature vectors collapse to subspaces spanned by principal components of features and weights, and the weight vector G…

Federated learning gets a speed boost: New extrapolation strategies significantly improve FedProx’s convergence, offering both theoretical backing and practical enhancements.

Poisson Midpoint Method quadratically accelerates Langevin Monte Carlo for diffusion models, achieving high-quality image generation with significantly fewer computations.