Episodic Reward

Episodic reward, where reinforcement learning agents receive feedback only at the end of a task, presents a significant challenge due to the delayed and sparse nature of the signal. Current research focuses on developing methods to decompose this episodic reward into more informative, step-wise proxy rewards, often employing attention mechanisms or novel reward structures within multi-agent and single-agent settings. This involves designing algorithms that efficiently utilize episodic memory and learn effective reward redistribution strategies, leading to improved sample efficiency and performance in various tasks, particularly in robotics and complex game environments. These advancements are crucial for enabling reinforcement learning agents to solve more challenging real-world problems with delayed feedback.

Papers

March 6, 2024

Bidirectional Progressive Neural Networks with Episodic Return Progress for Emergent Task Sequencing and Robotic Skill Transfer
Suzan Ece Ada, Hanne Say, Emre Ugur, Erhan Oztop
Neural Network Skill Transfer Robotic Skill Episodic Reward

March 2, 2024

Efficient Episodic Memory Utilization of Cooperative Multi-Agent Reinforcement Learning
Hyungho Na, Yunkyeong Seo, Il-chul Moon
Cooperative Multi Agent Reinforcement Learning Episodic Memory Episodic Control Episodic Reward

December 17, 2023

Episodic Return Decomposition by Difference of Implicitly Assigned Sub-Trajectory Reward
Haoxin Lin, Hongqiu Wu, Jiaji Zhang, Yihao Sun, Junyin Ye, Yang Yu
Qualitative Difference Proxy Reward Trajectory Wise Reward Episodic Reward Return Decomposition

January 12, 2022

Agent-Temporal Attention for Reward Redistribution in Episodic Multi-Agent Reinforcement Learning
Baicen Xiao, Bhaskar Ramasubramanian, Radha Poovendran
Multi Agent Reinforcement Learning Temporal Attention Reward Redistribution Episodic Reward

November 26, 2021

Learning Long-Term Reward Redistribution via Randomized Return Decomposition
Zhizhou Ren, Ruihan Guo, Yuan Zhou, Jian Peng
Reinforcement Learning Episodic Reinforcement Learning Proxy Reward Episodic Reward Reward Redistribution Return Decomposition

Episodic Reward

Papers

Bidirectional Progressive Neural Networks with Episodic Return Progress for Emergent Task Sequencing and Robotic Skill Transfer

Efficient Episodic Memory Utilization of Cooperative Multi-Agent Reinforcement Learning

Episodic Return Decomposition by Difference of Implicitly Assigned Sub-Trajectory Reward

Agent-Temporal Attention for Reward Redistribution in Episodic Multi-Agent Reinforcement Learning

Learning Long-Term Reward Redistribution via Randomized Return Decomposition