Finding Optimal Behaviors with Genetic Algorithms

Abstract

Optimality models used in behavioral ecology, and especially on insect parasitoids, have taken a variety of approaches, from classical analytical tools to individualbased simulations. The increasing awareness that much of the observable behavior in parasitoids depends on the state of the insect (be it the physiological or informational state) has led to the increasing use of stochastic dynamic programming models. However, optimal behaviors of one individual often depend upon the behavior of conspecifics, further complicating the issue. While classical game theory may be applied when behavior is not state dependent, genetic algorithms (GA) provide a powerful way in finding optimal behaviors for situations where such optimal behavior depends upon an animal's state and on the frequency of alternative behaviors of conspecifics. More generally, GAs can be used when there is a need to find the optimal strategy among a number of different alternative behaviors whose number is far too great to be exhaustively checked. GAs are search algorithms that proceed in a fashion analogous to natural selection. They are individual-based simulations that identify optimal solutions by searching the enormous space of potential solutions mimicking the process of evolution in biological systems.In this chapter, we discuss research questions in parasitoid behavioral ecology that might benefit from applying GA as a research tool for optimality models. Examples will be developed and explained. Among others, using the foraging problem of estimating habitat quality, we show how fast GAs find optimal solutions for parameter spaces that are impossible to solve numerically otherwise.

The combined effect of host and food availability on optimized parasitoid life‐history traits based on a three‐dimensional trade‐off surface

“…We used a genetic algorithm (Hoffmeister & Wajnberg, 2008; Ruxton & Beauchamp, 2008; Sumida, Houston, Mcnamara, & Hamilton, 1990) to compute the location on the 3D trade‐off surface providing the trade‐off between initial investment in energy reserves, the number of eggs produced, and the size of the eggs that maximizes the foraging female's fitness. Such a computer‐assisted optimization tool has been already used to solve ecological questions ( e.g.…”

“…Such a computer‐assisted optimization tool has been already used to solve ecological questions ( e.g. Sumida et al, 1990; Hoffmeister & Wajnberg, 2008; Ruxton & Beauchamp, 2008; Wajnberg, Coquillard, Vet, & Hoffmeister, 2012; Wajnberg, Hoffmeister, & Coquillard, 2012). In each environmental condition tested, the algorithm was run with 100 chromosomes (each with three loci for the traits being optimized), corresponding to 100 possible female genotypes.…”

Optimal Resource Allocation to Survival and Reproduction in Parasitic Wasps Foraging in Fragmented Habitats

“…In each environmental situation tested, values of the three parameters G1, G2 and G3 that maximize the reproductive success of the simulated animals were identified by means of a genetic algorithm [64] . Although remaining computationally simple, such a numerical method, that has been used to solve several other ecological questions ( e.g.…”

Evolutionary constraints on polyembryony in parasitic wasps: a simulation model

“…Parasitoids attain optimal fitness when their trait value is identical to host quality/type, namely when they are perfectly adapted to their host. Just like in genetic algorithms (Sumida et al 1990, Ruxton and Beauchamp 2008, Hoffmeister and Wajnberg 2008, the individuals at each generation that contribute to the next were chosen using a roulette wheel selection process (Bäck 1996), in which individuals with the higher fitness have a higher chance of being selected. This random process imitated natural selection, and enabled us to choose the mode of reproduction of all females and the genotype of all of their eggs.…”