A fast neural network simulator for state transition analysis

This paper describes an efficient simulator for state transition analysis o f multivalued continuous-time neural networks, where the multivalued transf er function of neuron is regarded as a stepwise constant one. Use of stepwi se constant method enables to analyse the state transition of the network w ithout solving explicitly the differential equations. This method also enab les to select the optimal timestep in numerical integration. The proposed m ethod is implemented on the simulator and applied to the general neural net work analysis. Furthermore, this is compared with the conventional simulato rs. Finally, it is shown that our simulator is drastically faster and more practical than the conventional simulators.