Proximal Curriculum with Task Correlations for Deep Reinforcement Learning

arXiv:2405.02481v1 Announce Type: new
Abstract: Curriculum design for reinforcement learning (RL) can speed up an agent's learning process and help it learn to perform well on complex tasks. However, existing techniques typically require domain-specific hyperparameter tuning, involve expensive optimization procedures for task selection, or are suitable only for specific learning objectives. In this work, we consider curriculum design in contextual multi-task settings where the agent's final performance is measured w.r.t. a target distribution over complex tasks. We base our curriculum …

abstract agent arxiv correlations cs.ai cs.lg curriculum design domain however hyperparameter learn optimization process reinforcement reinforcement learning speed tasks type