A pulse neural network reinforcement learning algorithm for partially observable Markov decision process

摘要

In this paper, we propose a new pulse neural network model and its reinforcement learning algorithm. The network is a feed-forward network with two hidden layers. The first hidden layer consists of pulse neurons with low decay rate of internal state, and the second layer consists of pulse neurons with high decay rate. The main purpose of this model is to utilize pulse neurons' ability for handling sequential input in partially observable Markov decision process. Its performance is confirmed by two kinds of computer simulations.