Reconstruction of Brain-inspired Visual Spiking Neural Network on BiCoSS

Abstract

At present, most of the neural networks, such as convolution neural network and brain-inspired spiking neural network (SNN), are realized by software simulation with the general computer with von Neumann architecture. More and more researchers use digital neuromorphic approach to implement SNN models for brain-inspired computing. In this study, we present a digital neuromorphic implementation on BiCoSS, and implement the visual pathway SNN model. This study focuses on the implementation of visual pathway using SNN model based on BiCoSS. Firstly, the architecture and design of the brain-inspired computing platform are described. Secondly, the hardware design and implementation of the SNN model for visual pathway are introduced. Then, the hardware implementation of self-stable LIF neurons and event-driven third-order spike timing-dependent plasticity (STDP) learning rule are introduced in detail. Finally, the hardware implementation results and functions of the whole model are verified. This study can be applied in a number of fields, including unmanned vehicles, autonomous robots, and internet of things (IoTs).