IR-VIC: Unsupervised Discovery of Sub-goals for Transfer in RL

IJCAI 2020

IR-VIC: Unsupervised Discovery of Sub-goals for Transfer in RL

Jan 20, 2021
|
32 views
|
Details
Abstract: We propose a novel framework to identify subgoals useful for exploration in sequential decision making tasks under partial observability. We utilize the variational intrinsic control framework (Gregor et.al., 2016) which maximizes empowerment – the ability to reliably reach a diverse set of states and show how to identify sub-goals as states with high necessary option information through an information theoretic regularizer. Despite being discovered without explicit goal supervision, our subgoals provide better exploration and sample complexity on challenging grid-world navigation tasks compared to supervised counterparts in prior work. Authors: Nirbhay Modhe, Prithvijit Chattopadhyay, Mohit Sharma, Abhishek Das, Devi Parikh, Dhruv Batra, Ramakrishna Vedantam (Georgia Institute of Technology, Facebook AI Research)

Comments
loading...