Machine Learning

Constrained Markov Decision Processes (CMDPs) get an improved sample complexity bound of Õ(1/ε) via a new algorithm, surpassing the existing O(1/ε²) bound.

Space partitioning dramatically speeds up relative entropy coding (REC) for neural compression, achieving 5-15% better bitrates than previous methods.

VQShape: a pre-trained model uses abstracted shapes as interpretable tokens for generalizable time-series classification, achieving comparable performance to black-box models and excelling in zero-sho…

STAR framework leverages state abstraction for consistent, low-variance off-policy evaluation in reinforcement learning, outperforming existing methods.

Absorb & Escape: a novel post-training sampling method that overcomes single model limitations by combining Autoregressive (AR) and Diffusion Models (DMs), generating high-quality heterogeneous genomi…

A2PO: A novel offline RL method tackles constraint conflicts in mixed-quality datasets by disentangling behavior policies with a conditional VAE and optimizing advantage-aware constraints, achieving s…

ATAC-Diff: A versatile diffusion model for high-quality single-cell ATAC-seq data generation and analysis, surpassing state-of-the-art.

New normative framework for decision confidence models diverse tasks by incorporating rewards, priors, and uncertainty, outperforming existing methods.

Unified pessimism principle in offline RL conquers data sparsity & model mismatch, achieving near-optimal performance across various divergence models.

A unified confidence sequence (CS) construction for generalized linear models (GLMs) achieves state-of-the-art regret bounds for contextual bandits, notably a poly(S)-free regret for logistic bandits.

New 2sED measure effectively bounds deep learning model complexity, correlating well with training error and offering efficient computation, particularly for deep models via a layerwise approach.

Trifle: Tractable inference for Offline RL achieves state-of-the-art results by using tractable generative models to overcome the inference-time suboptimality of existing sequence modeling approaches.

TACO, a novel topology-aware graph coarsening framework, tackles catastrophic forgetting in continual graph learning by efficiently preserving topological information during experience replay, signifi…

Bounded human feedback hinders large AI model training. This paper introduces hierarchical reinforcement learning to enable scalable oversight, efficiently acquiring feedback and learning optimal poli…

FedSAK, a novel federated multi-task learning framework, flexibly handles data, model, and task heterogeneity using tensor trace norm to learn correlations among client models, achieving superior perf…

Learn optimal device placement for neural networks with HSDAG, a novel framework boosting inference speed by up to 58.2%!

Single-step Sharpness-Aware Minimization (S2-SAM) achieves efficient and accurate sparse training by approximating sharpness perturbation via prior gradient information, incurring zero extra cost and …

SimGRL: A novel framework boosts visual reinforcement learning’s generalization by mitigating imbalanced saliency and observational overfitting through a feature-level frame stack and shifted random o…

New scalable algorithm reconstructs brain dynamics from short neuroimaging data, overcoming limitations of existing methods and enabling more accurate, efficient analysis of large-scale brain activity…

A novel machine learning recipe drastically accelerates charge density prediction in density functional theory, achieving state-of-the-art accuracy while being significantly faster than existing metho…