Using explicit averaging fitness for studying the behaviour of rats in a maze

Abstract

In this paper we study the performance of different evolutionary strategies based on explicit averaging. On a previous study, (Costa et al., 2012) proposed a probabilistic fitness function for an agent model based on neural networks and genetic algorithms employed to investigate the behaviour of rats in an elevated plus-maze (EPM). Differently from other computational models, the virtual rat proposed in (Costa et al., 2012) is not built based on experimental data comparisons with real rats, but, instead, is based on a behavioural model exploring the conflict between fear and anxiety. Despite the good results of the proposed agent, the effects of the uncertain fitness functions in the evolutionary learning process were not studied in the previous study. In our experiments we found significant differences in the performance of the genetic algorithm when the fitness of the individuals is sampled different times thus enabling us to define the best strategy for the studied problem.

Investigation of rat exploratory behavior via evolving artificial neural networks

“…There are many different strategies to deal with the optimization in uncertain environments. For the neuroevolutionary Model 1, some strategies are discussed by Costa et al (2013). The strategy adopted here is to evolve the virtual rats for a large number of generations.…”

“…The fitness function (Costa et al, 2012(Costa et al, , 2013 is composed by the weighted sum of a reward term (representing the curiosity of the rat) and a punishment term (representing its exposition to danger). The fitness function is given by:…”

“…Some studies have shown that, after this period, rat's interest in the maze is naturally decreased (Montgomery, 1955). M a n u s c r i p t for the investigation of the behavior of rats in the EPM (Salum et al, 2000;Giddings, 2002;Miranda et al, 2009;Shimo et al, 2010;Tejada et al, 2010;Costa et al, 2012Costa et al, , 2013. One of the advantages of computational models is their ability to predict real phenomena.…”

“…Investigating the best architectures for an ANN employed in the virtual rat can provide important insights about strategies used by the rats while navigating in the EPM. In the neuroevolutionary models (Costa et al, 2012(Costa et al, , 2013, the choice of the ANN's architecture is defined by the designer using heuristic rules and/or a trial and error approach. However, the number of possible architectures is huge, and, as a consequence, such approaches can result in unsatisfactory models.…”

“…Since 2012, we have been exploring neuroevolutionary models from a different perspective, where it is not necessary to employ data obtained from the trajectory of real rats to build the virtual rat (Costa et al, 2012(Costa et al, , 2013 In order to facilitate the study of rodent trajectories in the maze, the EPM arms are divided into equal positions (rectangles).…”