ΑΙhub.org



BAIR blog


website   |   @berkeley_ai   |  

The Berkeley Artificial Intelligence Research (BAIR) Lab brings together UC Berkeley researchers across the areas of computer vision, machine learning, natural language processing, planning, and robotics. BAIR includes over two dozen faculty and more than a hundred graduate students pursuing research on fundamental advances in the above areas as well as cross-cutting themes including multi-modal deep learning, human-compatible AI, and connecting AI with other scientific disciplines and the humanities. The BAIR Blog provides an accessible, general-audience medium for BAIR researchers to communicate research findings, perspectives on the field, and various updates. Posts are written by students, post-docs, and faculty in BAIR, and are intended to provide relevant and timely discussion of research findings and results, both to experts and the general audience.




recent posts:


›   Should I use offline RL or imitation learning?


›   Offline RL made easier: no TD learning, advantage reweighting, or transformers


›   Unsupervised skill discovery with contrastive intrinsic control


›   imodels: leveraging the unreasonable effectiveness of rules


›   The unsupervised reinforcement learning benchmark


›   Sequence modeling solutions for reinforcement learning problems


›   Which mutual information representation learning objectives are sufficient for control?


›   Bridge data: boosting generalization of robotic skills with cross-domain datasets


›   Why generalization in RL is difficult: epistemic POMDPs and implicit partial observability


›   Designs from data: offline black-box optimization via conservative training


›   A first-principles theory of neural network generalization


›   Making RL tractable by learning more informative reward functions: example-based control, meta-learning, and normalized maximum likelihood


›   PICO: Pragmatic compression for human-in-the-loop decision-making


›   Unsolved ML safety problems


›   Distilling neural networks into wavelet models using interpretations


›   What can I do here? Learning new skills by imagining visual affordances


›   Universal weakly supervised segmentation by pixel-to-segment contrastive learning


›   The surprising effectiveness of PPO in cooperative multi-agent games


›   Learning what to do by simulating the past


›   The importance of hyperparameter optimization for model-based reinforcement learning


›   Pretrained transformers as universal computation engines


›   Maximum entropy RL (provably) solves some robust RL problems


›   Self-supervised policy adaptation during deployment


›   The successor representation, gamma-models, and infinite-horizon prediction


›   Does GPT-2 know your phone number?


›   Offline reinforcement learning: how conservative algorithms can enable new applications


›   Learning state abstractions for long-horizon planning


›   EvolveGraph: dynamic neural relational reasoning for interacting systems


›   Training on test inputs with amortized conditional normalized maximum likelihood


›   Goodhart’s law, diversity and a series of seemingly unrelated toy problems





next page →




©2021 - Association for the Understanding of Artificial Intelligence


 












©2021 - ROBOTS Association