Deep Reinforcement Learning with Permutation Equivariant Neural Network for Multi-Armed Bandit
- 10 February 2026
- journal article
- Published by World Scientific Pub Co Pte Ltd in International Journal on Artificial Intelligence Tools
- Vol. 35 (02)
Abstract
Permutation equivariance (PE) is a property widely present in mathematics and machine learning. Classic deep reinforcement learning (DRL) algorithms, such as Deep Q-Network (DQN), require thorough exploration of the state space to achieve optimal performance. For a PE problem such as the Multi-Armed Bandit (MAB), the PE property helps reduce the space that needs to be explored. This paper proposes PEDQN, a PE DRL framework based on DQN that applies a PE neural network structure. Our MAB experiments show that PEDQN has clear advantages over DQN with a fully connected network and achieves the same or better performance than UCB1 when tested in the same environment as the training. Additionally, PE-based architectures can adapt to varying input sizes, with pretrained models accelerating learning in new environments.Keywords
Funding Information
- French National Research Agency (ANR-21-CE48-0003)
- “France 2030” government investment plan managed by ANR (ANR-23-PEIA-0006)
This publication has 11 references indexed in Scilit:
- Permutation Equivariant Deep Reinforcement Learning for Multi-Armed Bandit2024
- ATMoS+: Generalizable Threat Mitigation in SDN Using Permutation Equivariant and Invariant Deep Reinforcement LearningIEEE Communications Magazine, 2022
- Geometric Invariant Representation Learning for 3D Point Cloud2021
- Historical Best Q-Networks for Deep Reinforcement Learning2018
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation2017
- Deep Reinforcement Learning with Double Q-LearningProceedings of the AAAI Conference on Artificial Intelligence, 2016
- Human-level control through deep reinforcement learningNature, 2015
- Simple Artificial Neural Networks That Match Probability and Exploit and Explore When Confronting a Multiarmed BanditIEEE Transactions on Neural Networks, 2009
- Finite-time Analysis of the Multiarmed Bandit ProblemMachine Learning, 2002
- Q-Learning for Bandit Problems1995