Convergence of a Human-in-the-Loop Policy-Gradient Algorithm With Eligibility Trace Under Reward, Policy, and Advantage Feedback

Abstract

Fluid human-agent communication is essential for the future of human-in-the-loop reinforcement learning. An agent must respond appropriately to feedback from its human trainer even before they have significant experience working together. Therefore, it is important that learning agents respond well to various feedback schemes human trainers are likely to provide. This work analyzes the COnvergent Actor-Critic by Humans (COACH) algorithm under three different types of feedback-policy feedback, reward feedback, and advantage feedback. For these three feedback types, we find that COACH can behave sub-optimally. We propose a variant of COACH, episodic COACH (E-COACH), which we prove converges for all three types. We compare our COACH variant with two other reinforcement-learning algorithms: Q-learning and TAMER.

arXiv commentAccepted into ICML 2021 workshops Human-AI Collaboration in Sequential Decision-Making and Human in the Loop Learning

References in corpus (5)

• How to Escape Saddle Points Efficiently
• On the Theory of Policy Gradient Methods: Optimality, Approximation, and Distribution Shift
• Interactive Learning from Policy-Dependent Human Feedback
• Policy Gradient Methods for Reinforcement Learning with Function Approximation and Action-Dependent Baselines
• Local limit theorems for occupancy models