Evolutionary Multiobjective Optimization of Spiking Neural Networks

The optimization of spiking neural networks is a difficult task and has been handled previously by evolutionary algorithms. The usage of evolutionary algorithms allows for the optimization of more complex problem types, including multiobjective optimization. In this work, we present a method for multiobjective neuroevolution of spiking neural networks. The underlying multiobjective approach is shown to be highly effective on a suite of test problems, and the complete neuroevolution algorithm is applied to the optimization of an expanded classical control problem. The algorithm is found to be capable of evolving a diverse set of highly capable solutions, even on configurations with a fairly high number of objectives.