Genetic-Gated Networks for Deep Reinforcement

TL;DR

This paper introduces Genetic-Gated Networks (G2Ns), which combine genetic algorithms and gradient-based methods to improve reinforcement learning by enhancing sample efficiency and performance.

Contribution

The paper presents a novel neural network architecture that integrates genetic gating with traditional training methods for reinforcement learning.

Findings

G2Ns significantly improve sample efficiency in reinforcement learning tasks.

G2Ns outperform baseline models in various reinforcement learning benchmarks.

Multiple models can be constructed simultaneously using different chromosomes.

Abstract

We introduce the Genetic-Gated Networks (G2Ns), simple neural networks that combine a gate vector composed of binary genetic genes in the hidden layer(s) of networks. Our method can take both advantages of gradient-free optimization and gradient-based optimization methods, of which the former is effective for problems with multiple local minima, while the latter can quickly find local minima. In addition, multiple chromosomes can define different models, making it easy to construct multiple models and can be effectively applied to problems that require multiple models. We show that this G2N can be applied to typical reinforcement learning algorithms to achieve a large improvement in sample efficiency and performance.