A Spiking Neural Networks with Probability Information Transmission

Abstract The spiking neural network provides a potential computing paradigm for simulating the complex information processing mechanism of the brain. Even though there are many theoretical and practical achievements, several crucial problems remain to be addressed for the existing spiking learning algorithm. In this paper, a Probabilistic Spike Response Model (PSRM), of which ignition mode is determined neither by the difference between the threshold and membrane voltage nor in the form of pulses, is proposed from a probabilistic perspective. The PSRM reconstructs a probability relationship between the membrane voltage and neuron ignition to transmit information in the form of probabilities and redefines pulses. The expression of the pulse sequence makes the PSRM model continuous and differentiable so as to avoid the difficulty in using supervised learning algorithms. In our study, the single-layer learning algorithm and multilayer learning algorithm based on the PSRM model are also given. As shown in our experiments, the proposed method is of theoretical and practical value.