Adaptation in Natural and Artificial Systems
Holland described a search in which a population of solutions crosses over, mutates and is selected by fitness.
Why it matters
Optimisation gained a method that needs neither a derivative nor a model of the problem, only a way to score a result.
The book introduces schemata and the schema theorem: selection grows the share of short successful fragments of a solution exponentially. Unlike gradient methods it needs no differentiability, so it works on discrete and discontinuous problems. Evolutionary strategies and neuroevolution later grew from this line.