Optimization of Neural Networks
The operation of a spiking neural network is controlled by a large number of parameters. The relationship between the network parameters and network output is highly non-linear and is intractable to mathematical analysis.
The aim of this project is to devise methods for tuning network parameters to obtain a desired network behaviour. We use a genetic breeder algorithm with an adaptive mutation rate to search for the optimal parameters. A staged evolution approach is used to increase the rate of convergence of the algorithm.