Genetic Algorithm — Fitness Evolution (2D)

Parameters

Population size 80 Mutation rate 0.05 Crossover rate 0.80 Selection method

0generation

—best fitness

—avg fitness

—alive

Each individual is a 20-gene real-valued genome. Fitness = 1 − mean-squared error against a fixed random target vector. Parents are chosen by tournament or roulette-wheel selection, combined by uniform crossover, and perturbed by Gaussian mutation. The single best genome survives unchanged each generation (elitism), so best fitness never regresses. Top chart: best/average fitness per generation. Bottom chart: current population's fitness distribution.