Double DQN

Double Deep Q-Networks (Double DQN) is a reinforcement learning algorithm aiming to improve the stability and performance of Q-learning by mitigating overestimation bias. Current research focuses on enhancing Double DQN's efficiency and applicability through techniques like incorporating ensemble methods, leveraging problem structure (e.g., weakly coupled systems), and integrating heuristic guidance or local planning to improve sample efficiency. These advancements are impacting various fields, including traffic signal control and combinatorial optimization, by enabling the development of more robust and effective AI agents for complex decision-making problems.

Papers

October 28, 2023

Weakly Coupled Deep Q-Networks
Ibrahim El Shar, Daniel R. Jiang
Deep Reinforcement Learning Q Learning Deep Q Network DQN Agent Double DQN

June 17, 2023

Vanishing Bias Heuristic-guided Reinforcement Learning Algorithm
Qinru Li, Hao Xiang
Reinforcement Learning Policy Gradient Q Learning DQN Agent Double DQN

January 29, 2023

Sample Efficient Deep Reinforcement Learning via Local Planning
Dong Yin, Sridhar Thiagarajan, Nevena Lazic, Nived Rajaraman, Botao Hao, Csaba Szepesvari
Reinforcement Learning Uncertainty Aware Sample Efficient Reinforcement Learning Local Planning Double DQN Initial State Distribution

July 18, 2022

A Deep Reinforcement Learning Approach for Finding Non-Exploitable Strategies in Two-Player Atari Games
Zihan Ding, Dijia Su, Qinghua Liu, Chi Jin
Markov Game Deep Reinforcement Learning Approach Tabular Markov Decision Process ATARI Game Nash Equilibrium Policy Double DQN

March 10, 2022

Random Ensemble Reinforcement Learning for Traffic Signal Control
Ruijie Qi, Jianbin Huang, He Li, Qinglin Tan, Longji Huang, Jiangtao Cui
Traffic Signal Control Ensemble Deep Reinforcement Learning Double DQN

January 17, 2022

An Improved Reinforcement Learning Algorithm for Learning to Branch
Qingyu Qu, Xijun Li, Yunfan Zhou, Jia Zeng, Mingxuan Yuan, Jie Wang, Jinhu Lv, Kexin Liu, Kun Mao
LeArning Abstract Offline Reinforcement Learning Multi Channel Branch Branch and Cut Double DQN

August 9, 2021

Modified Double DQN: addressing stability
Shervin Halat, Mohammad Mehdi Ebadzadeh, Kiana Amani
Core Stability DQN Agent Double DQN