Solving the Spike Feature Information Vanishing Problem in Spiking Deep Q Network with Potential Based Normalization

TL;DR

This paper introduces a potential based layer normalization method to directly train deep spiking Q networks, effectively addressing the spike feature vanishing problem and improving performance on Atari games.

Contribution

It proposes a novel potential based layer normalization technique for training deep spiking Q networks directly, overcoming feature vanishing issues in reinforcement learning.

Findings

PL-SDQN outperforms ANN-SNN conversion methods.

The method effectively mitigates spike feature vanishing.

Improved performance on Atari game benchmarks.

Abstract

Brain inspired spiking neural networks (SNNs) have been successfully applied to many pattern recognition domains. The SNNs based deep structure have achieved considerable results in perceptual tasks, such as image classification, target detection. However, the application of deep SNNs in reinforcement learning (RL) tasks is still a problem to be explored. Although there have been previous studies on the combination of SNNs and RL, most of them focus on robotic control problems with shallow networks or using ANN-SNN conversion method to implement spiking deep Q Network (SDQN). In this work, we mathematically analyzed the problem of the disappearance of spiking signal features in SDQN and proposed a potential based layer normalization(pbLN) method to directly train spiking deep Q networks. Experiment shows that compared with state-of-art ANN-SNN conversion method and other SDQN works, the proposed pbLN spiking deep Q networks (PL-SDQN) achieved better performance on Atari game tasks.