Sound localization through evolutionary learning applied to spiking neural networks

A biologically based learning framework is established to study neural modeling with respect to sound source localization. This involves a 2-dimensional environment wherein agents must locate sound sources that are periodically resituated whilst emitting pulses at regular intervals. Agents employ a spiking neural model that controls movement on the basis of binaural inputs, and evolutionary learning (EL) is applied to evolve neural connectivity and weights. It is demonstrated that agents are successfully able to locate sound sources and that the simulative framework can be extended to address questions pertaining to the evolution of spiking neural networks.