Learning to Model What Matters // Model-Based Reinforcement Learning скачать в хорошем качестве

Learning to Model What Matters // Model-Based Reinforcement Learning

Today's paper: Goal-Aware Prediction: Learning to Model What Matters Abstract ### "Learned dynamics models combined with both planning and policy learning algorithms have shown promise in enabling artificial agents to learn to perform many diverse tasks with limited supervision. However, one of the fundamental challenges in using a learned forward dynamics model is the mismatch between the objective of the learned model (future state reconstruction), and that of the downstream planner or policy (completing a specified task). This issue is exacerbated by vision-based control tasks in diverse real-world environments, where the complexity of the real world dwarfs model capacity. In this paper, we propose to direct prediction towards task relevant information, enabling the model to be aware of the current task and encouraging it to only model relevant quantities of the state space, resulting in a learning objective that more closely matches the downstream task. Further, we do so in an entirely self-supervised manner, without the need for a reward function or image labels. We find that our method more effectively models the relevant parts of the scene conditioned on the goal, and as a result outperforms standard task-agnostic dynamics models and model-free reinforcement learning." 0:00 Problems with most world-model learning approaches 0:28 Don't Learn Everything 01:21 What is Useful 01:40 Paper of today 02:35 Usual WM architecture 03:37 GAP: Learn What Matters- Architecture 05:00 Advantages of the Proposed Solution 06:00 Hindsight Relabeling/Goal Relabeling 07:30 Does it work? Results #reinforcementlearning #GAP #modelbased #deeplearning #learningwhatmatters #hindsightrelabeling