Fitnessless coevolution

Abstract

We introduce fitnessless coevolution (FC), a novel method of comparative one-population coevolution. FC plays games between individuals to settle tournaments in the selection phase and skips the typical phase of evaluation. The selection operator applies a single-elimination tournament to a randomly drawn group of individuals, and the winner of the final round becomes the result of selection. Therefore, FC does not involve explicit fitness measure. We prove that, under a condition of transitivity of the payoff matrix, the dynamics of FC is identical to that of the traditional evolutionary algorithm. The experimental results, obtained on a diversified group of problems, demonstrate that FC is able to produce solutions that are equally good or better than solutions obtained using fitness-based one-population coevolution with different selection methods.

Evolving Tic-Tac-Toe Playing Algorithms Using Co-Evolution, Interactive Fitness and Genetic Programming

“…The Function Set contains the following game operations: AND, OR, IF, MINE, YOURS, OPEN, and PLAY-AT. These functions return either a position on the game board or the value NIL, which is introduced in [8] and discussed shortly. The "IF" function is the only operator in the function set that requires three arguments.…”

“…According to [8] one population Co-Evolution is computationally more feasible than a two population Co-Evolution, therefore, we utilized Co-Evolution with a single population.…”

“…[7] introduced a Co-Evolutionary system in which they utilized Shared Sampling and Competitive Fitness Sharing. [8] performed an interesting implementation using single population Co-Evolution. The authors of [8] have introduced Fitness-less Co-Evolution in order to generate Tic-Tac-Toe playing algorithm.…”

“…[8] performed an interesting implementation using single population Co-Evolution. The authors of [8] have introduced Fitness-less Co-Evolution in order to generate Tic-Tac-Toe playing algorithm. However, they did not utilize Mutation operation along with Reproduction and Cross-over.…”

Evolving strategy for a probabilistic game of imperfect information using genetic programming

“…This observation led us to propose the approach of fitnessless coevolution. The key idea is to combine the typical evaluation and selection phases of evolutionary algorithm into one step of fitnessless selection [12]. Technically, we skip the evaluation and proceed directly to selection, which works like a single-elimination tournament played between k individuals randomly drawn from the population.…”