Design And Implementation Of Spiking Neural Network Platform For Darwin Chip

Brain-like Computing is becoming research focus due to the power consumption and heat radiation problem of Von Neumann based computer architecture.The basic idea of Brain-like Computing is using the conception of biological neural network to design computer systems,and then improve the performance and reduce power consumption over the specific applications.Spiking neural network is developed under the simulation of biological neural network,which make it becoming the main research direction of Brain-like computing.There are many break throughs over the design of Brain-like Chip,and Darwin is one of Brain-like Chips.Darwin is a coprocessor chip designed by ZheJiang University and Hangzhou Dianzi University.We need to use specific application systems to validate and evaluate our Darwin chip,and we should built a appropriate spiking neural network for Darwin chip.So this paper designed and implemented a spiking neural network platform,which provided a customizable and intuitive platform for application design of Brain-like computing chip.And we finally implemented a real-time demonstration system for Darwin chip using our platform.The main contributions of this paper are:1)Designed and implemented a spiking neural network for Darwin chip,and imple-mented the algorithm of converting artificial neural network to spiking neural network.2)Built and implemented a spiking neural network simulation platform,which can config the network structure and parameters by visual interface.The platform can also implement spiking neural networks with customized structure easily,and visualize the pro-cedure and results of spiking neural network simulation.3)Designed and implemented a real-time demonstration system of control virtual bas-ketball moves by imagination for Darwin chip using our spiking neural network simulation platform.