evolutionary optimization

The main bottleneck in evolutionary algorithms is always the simulation (expressivity, complexity, open-endedness, …).

Evolution Strategies

Evolution Strategies (ES) is a class of black box optimization algorithms [Rechenberg and Eigen, 1973, Schwefel, 1977] that are heuristic search procedures inspired by natural evolution: At every iteration (“generation”), a population of parameter vectors (“genotypes”) is perturbed (“mutated”) and their objective function value (“fitness”) is evaluated. The highest scoring parameter vectors are then recombined to form the population for the next generation, and this procedure is iterated until the objective is fully optimized. Algorithms in this class differ in how they represent the population and how they perform mutation and recombination.

Link to original

ES: $(\mu +\lambda )$, $(\mu ,\lambda )$

Generally: Generate population → Evaluate → Keep best → Generate population (based on best) → …
$\mu$ … elites, best performing individuals of a generation
$\lambda$ … population size, essentiall
$(\mu +\lambda )$ … give the parents a chance to compete with their children
$(\mu ,\lambda )$ … replace parents by children (can help overcome local optima → advantage over gradient-based methods; bio inspo: death of old age)

NEAT
cma-es
zero order optimization