evolutionary algorithm
A gradient-free optimization method inspired by biological evolution. It searches for optimal solutions by maintaining a population of candidate solutions, generating new individuals through mutation and crossover, and retaining better individuals according to a fitness function. It is suited to discrete or non-convex search spaces.
Timeline (last 90 days)
Evolutionary algorithm setup: each generation tries 64 mutated copies and keeps the best-performing 16.
The statement in the article that "each generation tries 64 mutated copies and keeps the best 16" is a typical (1+λ) or truncation selection strategy, which is gradient-free optimization.
Validating evolutionary strategies in a simple simulated environment is low-cost and highly visualizable, making it suitable for exploring the intersection of collective behavior and neural-network decision-making.
A developer shared an experiment on Reddit: using an evolutionary algorithm to train tiny neural networks, getting 60 simulated emperor penguins to learn to huddle for warmth in a -40°C environment.
The generation-0 neural network did nothing and scored 38%; generation 2 surpassed the hand-crafted rules; after 41 generations of evolution, the final score reached 65%.
This "generate-evaluate-select" mechanism is isomorphic to the loop in evolutionary algorithms and is also often called best-of-N sampling or self-play-style improvement.
Evolutionary algorithms in spacecraft trajectory design can search for Pareto-optimal frontiers among multiple constraints, and ESA used evolutionary algorithms in the SMART-1 lunar probe mission to search for optimal transfer orbits.
All Facts (7)
Evolutionary algorithm setup: each generation tries 64 mutated copies and keeps the best-performing 16.
50%UnverifiedThe generation-0 neural network did nothing and scored 38%; generation 2 surpassed the hand-crafted rules; after 41 generations of evolution, the final score reached 65%.
50%UnverifiedThe statement in the article that "each generation tries 64 mutated copies and keeps the best 16" is a typical (1+λ) or truncation selection strategy, which is gradient-free optimization.
50%UnverifiedValidating evolutionary strategies in a simple simulated environment is low-cost and highly visualizable, making it suitable for exploring the intersection of collective behavior and neural-network decision-making.
50%UnverifiedThis "generate-evaluate-select" mechanism is isomorphic to the loop in evolutionary algorithms and is also often called best-of-N sampling or self-play-style improvement.
50%UnverifiedEvolutionary algorithms in spacecraft trajectory design can search for Pareto-optimal frontiers among multiple constraints, and ESA used evolutionary algorithms in the SMART-1 lunar probe mission to search for optimal transfer orbits.
50%UnverifiedA developer shared an experiment on Reddit: using an evolutionary algorithm to train tiny neural networks, getting 60 simulated emperor penguins to learn to huddle for warmth in a -40°C environment.
50%