Genetic Algorithm: Fitness Landscape Explorer

Evolution

Generation speed 2 / s

Landscape

Genetic Operators

Mutation rate (per gene) 15%

Mutation strength σ 0.60

Tournament size (selection pressure) 3

Population Stats

Generation

Best fitness

0.0%

Mean fitness

0.0%

Diversity σ

0.00

How it works

Each individual's genotype is a real-valued pair (x, y). The fitness landscape is a sum of Gaussian "hills" — a common way to model a rugged, multi-optimum search space:

f(x,y) = Σ A_i · exp( -((x-cx_i)² + (y-cy_i)²) / (2w_i²) )

Every generation the algorithm applies the three classic genetic operators:

Selection — tournament selection: pick k random individuals, keep the fittest. Larger k = stronger selection pressure = faster but less diverse.
Crossover — blend (arithmetic) crossover: child = p₁ + α(p₂ - p₁) for a random α ∈ [0,1] per gene, combining two parents' coordinates.
Mutation — with probability p per gene, add Gaussian noise N(0, σ²) to explore nearby terrain. The top 2 individuals are carried over unchanged (elitism), so the best solution found is never lost.

Low mutation + high selection pressure converges fast but often gets trapped climbing the nearest local peak instead of the global optimum — the classic exploration/exploitation trade-off in evolutionary search. Increase mutation strength or lower tournament size to keep the population exploring; watch the diversity σ (spread of the population in genotype space) collapse as everyone converges on one peak.

Gen 0

Individual (color = fitness) Best-so-far marker

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.