An information-theoretic landscape analysis of neuro-controlled embodied organisms

Abstract

Recently, there has been a lot of interest in evolving controllers for both physically simulated creatures as well as for real physical robots. However, a range of different ANN architectures are used for controller evolution, and, in the majority of the work conducted, the choice of the architecture used is made arbitrarily. No fitness landscape analysis was provided for the underlying fitness landscape of the controller's search space. As such, the literature remains largely inconclusive as to which ANN architecture provides the most efficient and effective space for searching the range of possible controllers through evolutionary methods. This represents the motivation for this paper where we compare the search space for four different types of ANN architecture for controller evolution through an information-theoretic analysis of the fitness landscape associated with each type of architecture.

On The Control Of Opinion In Social Networks

“…A landscape with many local optima or a high epistasis is said to be rugged, while one with fewer optima is relatively smooth [113,120]. A rugged landscape is typically considered to be more difficult to search as similar steps in the landscape have less correlation in fitness.…”

“…It is shown in [123] that these measures correlate with the performance of EA using various mutation and crossover operators. To allow the application of these landscape analysis measures, the fitness space is commonly assumed to be statistically isotropic [50,121,128], this is the case in related work with ENNs as well [114,120,122]. For a detailed review of a variety of fitness landscape measures see [91].…”