GAN-based Intrinsic Exploration For Sample Efficient Reinforcement Learning

• URL: http://arxiv.org/abs/2206.14256v1
• Date: Tue, 28 Jun 2022 19:16:52 GMT
• Title: GAN-based Intrinsic Exploration For Sample Efficient Reinforcement Learning
• Authors: Do\u{g}ay Kamar (1), Naz{\i}m Kemal \"Ure (1 and 2), G\"ozde \"Unal (1 and 2) ((1) Faculty of Computer and Informatics, Istanbul Technical University (2) Artificial Intelligence and Data Science Research Center, Istanbul Technical University)
• Abstract summary: We propose a Geneversarative Adversarial Network-based Intrinsic Reward Module that learns the distribution of the observed states and sends an intrinsic reward that is computed as high for states that are out of distribution. We evaluate our approach in Super Mario Bros for a no reward setting and in Montezuma's Revenge for a sparse reward setting and show that our approach is indeed capable of exploring efficiently.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: In this study, we address the problem of efficient exploration in reinforcement learning. Most common exploration approaches depend on random action selection, however these approaches do not work well in environments with sparse or no rewards. We propose Generative Adversarial Network-based Intrinsic Reward Module that learns the distribution of the observed states and sends an intrinsic reward that is computed as high for states that are out of distribution, in order to lead agent to unexplored states. We evaluate our approach in Super Mario Bros for a no reward setting and in Montezuma's Revenge for a sparse reward setting and show that our approach is indeed capable of exploring efficiently. We discuss a few weaknesses and conclude by discussing future works.

Related papers

Err

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.