Assessment and Comparison of Evolutionary Algorithms for Tuning a Bio-Inspired Retinal Model

摘要

One of the basic questions in neuroscience is how visual information is encoded in the retina. To design artificial retinal systems it is essential to emulate the mammalian retinal behaviour as well as possible. Furthermore, this is a question of primary interest in the design of an artificial neuroprosthesis where it is necessary to mimic the retina as much as possible. This work selects the best algorithm from a set of well-known evolutionary algorithms to perform a reliable tuning of a retinal model. The proposed design scheme optimizes various parameters belonging to different domains (that is, spatio-temporal filtering and neuromorphic encoding) to compare the biological and the simulated registers. Five algorithms have been tested: three different Genetic Algorithms (SPEA2, NSGA-II and NSGA-III), a Particle Swarm Optimization algorithm and a Differential Evolution algorithm. Their performances have been compared by using the hypervolume indicator.