📂Optimization

What is Sexual Selection in Genetic Algorithms?

Terminology ¹

In the genetic algorithm concept of crossover/recombination, a technique that increases diversity by placing a mating tag on part of a chromosome to impose constraints on mating partners is called sexual selection.

Description

Sexual selection can be implemented, for example, as follows: given a fitness function $f : X \to \mathbb{R}$, in a chromosome ▷eq2◯ the first index can indicate sex via $\left\{ 0, 1 \right\}$, and a function $g : X \to \mathbb{R}$ can be assigned to determine attractiveness. Natural selection is determined by fitness according to $f$, but which partners they meet is determined by $g$.

This can be seen as a proximal operator (프락시멀 오퍼레이터) implemented in the genetic-algorithm paradigm: maximizing fitness while moving somewhat closer to the value favored by $g$. In other words, the population adapts well to the environment $f$ while evolution is steered toward the traits I prefer according to the aesthetic criterion $g$.

Analogous to the concept of sexual selection in nature, sexual selection in genetic algorithms also complicates mating patterns and thereby enhances diversity. Rather than merely adapting to the environment and maximizing fitness, it promotes intra-sex competition and increases the probability that, even if immediate fitness is slightly reduced, crossover will produce superior offspring in subsequent generations.