Evolutionary algorithm

A population-based metaheuristic that uses Darwinian mechanisms like mutation and crossover to solve complex global optimization problems.

Evolutionary algorithms (EAs) mimic biological selection to navigate vast search spaces where traditional calculus-based methods fail. By maintaining a population of candidate solutions (phenotypes), the system applies stochastic operators—specifically mutation, recombination, and tournament selection—to iteratively refine performance. This methodology powers critical engineering feats, from NASA's ST5 spacecraft antenna design to high-frequency trading strategies and deep neural architecture search (NAS). EAs excel in multi-objective optimization (MOO), allowing stakeholders to balance competing constraints like cost and structural integrity without requiring a differentiable objective function.

https://www.nature.com/articles/s42256-020-00278-y