🧬
Neuroevolution Playground
Evolve neural networks via genetic algorithms. Watch populations learn to balance poles, fly through pipes, and navigate mazes.
Balance a pole on a moving cart
Speed:
Generation: 0
Best Fitness: 0.0
Avg Fitness: 0.0
Population: 100
Topology: 4-8-2
Evolution Parameters
Population Size100
Mutation Rate10%
Mutation Strength0.30
Crossover Rate70%
Elitism Count2
Tournament Size5
About This Task
A neural network controls a cart to balance a pole. Inputs: cart position, velocity, pole angle, angular velocity. Outputs: push left or right.
Fitness = steps survived + bonuses for staying centered and upright. Max fitness ~1020 (1000 steps + bonuses).
Neural networks and GA from scratch. Tournament selection, uniform crossover, Gaussian mutation. All computation in your browser.