Machine Learning

Model Sensitivity Aware Continual Learning (MACL) tackles the CL challenge by optimizing model performance based on parameter distribution, achieving superior old knowledge retention and new task perf…

Model LEGO (MDA) revolutionizes deep learning by enabling the creation of new models by assembling and disassembling task-aware components from pre-trained models, eliminating the need for retraining.

New computational method infers complex brain learning rules from experimental data, revealing active forgetting in reward learning.

Link-MoE boosts link prediction accuracy by strategically selecting the best model for each node pair, surpassing single-model approaches.

Non-linear Residual Gates (NoRGa) boosts prompt-based continual learning by theoretically framing prefix tuning as adding new experts to a pre-trained Mixture-of-Experts model, achieving state-of-the-…

New metric, λ-discrepancy, precisely detects & mitigates partial observability in sequential decision processes, significantly boosting reinforcement learning agent performance.

DrilDICE robustly tackles covariate shift in offline imitation learning by using a stationary distribution correction and a distributionally robust objective, significantly improving performance.

Proactive Defensive Backdoor (PDB) thwarts malicious backdoors by injecting a hidden defensive backdoor during training, suppressing attacks while maintaining model utility.

MiSO: a novel closed-loop brain stimulation framework optimizes stimulation parameters to achieve desired neural population activity states, overcoming limitations of current methods by merging data a…

MinUCB and LA-MinUCB, novel local Bayesian optimization algorithms, replace gradient descent with UCB minimization for efficient, theoretically-sound local search.

Minimax-optimal, computationally efficient algorithms are proposed for distributionally robust offline reinforcement learning, addressing challenges posed by function approximation and model uncertain…

IMAX-PPO: A novel multi-agent RL algorithm leveraging imitation learning to predict opponent actions, achieving superior performance in complex games.

METRIC FLOW MATCHING (MFM) generates smooth interpolations on data manifolds by minimizing kinetic energy, outperforming Euclidean methods and achieving state-of-the-art results in single-cell traject…

MetaCURL: First algorithm for non-stationary Concave Utility Reinforcement Learning (CURL), achieving near-optimal dynamic regret by using a meta-algorithm and sleeping experts framework.

Provable near-optimality in meta-RL is achieved using a novel bilevel optimization framework and universal policy adaptation algorithm.

MetaNCoV: Learn data-driven priors via non-injective change of variables for enhanced few-shot learning.

Meta-DT: Offline meta-RL masters unseen tasks via conditional sequence modeling and world model disentanglement, showcasing superior few-shot and zero-shot generalization.

Meta-Controller: A novel few-shot behavior cloning framework enables robots to generalize to unseen embodiments and tasks using only a few reward-free demonstrations, showcasing superior few-shot gene…

FG²U: a novel memory-efficient algorithm for unbiased stochastic approximation of meta-gradients in large-scale bi-level optimization, showing superior performance across diverse tasks.

MADPO, a novel MARL framework, uses mutual policy divergence maximization with conditional Cauchy-Schwarz divergence to enhance exploration and agent heterogeneity in sequential updating, outperformin…