Evolutionary Computation
Optimization algorithms inspired by biological evolution: selection, mutation, crossover. Includes genetic algorithms, swarm intelligence, and neuroevolution.
Sub-topics
John Holland's 1975 optimization method inspired by natural selection. Solutions evolve through selection, crossover, and mutation over generations.
Evolving computer programs represented as tree structures. John Koza (1992) pioneered evolving LISP programs to solve symbolic regression and classification problems.
Evolving neural network architectures and weights. NEAT (Stanley, 2002) co-evolves topology and weights. OpenAI showed evolution can match RL for policy search (2017).
Collective behavior of decentralized agents. Ant Colony Optimization (Dorigo, 1992) and Particle Swarm Optimization (Kennedy & Eberhart, 1995) solve combinatorial problems.
Black-box optimization using population-based search with Gaussian perturbations. CMA-ES (Hansen, 2001) is the gold standard. Competitive with RL for policy optimization.