Training framework based on multi model competition for deep reinforcement learning
Yixian lin · Journal of Physics Conference Series · 2021
Abstract In this paper, we propose a simple, effective and universal deep reinforcement learning training framework, inspired by A3C algorithm and genetic algorithm. The framework uses multi-process technology to realize multiple model competition and optimal evolution to optimize the deep neural network during the training process. The experimental results show that the proposed training framework can improve the training effect of reinforcement learning algorithms to a certain extent.