Week 11-12: Evolutionary Strategies & Genetic Algorithm

Evolution strategies are evolutionary algorithms that date back to the 1960s and that are most commonlyapplied to black-box optimization problems in continuous search spaces. Inspired by biological evolution,their original formulation is based on the application of mutation, recombination and selection in populationsof candidate solutions. From the algorithmic viewpoint, evolution strategies are optimization methodsthat sample new candidate solutions stochastically, most commonly from a multivariate normal probabilitydistribution. Their two most prominent design principles are unbiasedness and adaptive control of parametersof the sample distribution. In this overview the important concepts of success based step-size control, self-adaptation and derandomization are covered, as well as more recent developments like covariance matrixadaptation and natural evolution strategies. The latter give new insights into the fundamental mathematicalrationale behind evolution strategies. A broad discussion of theoretical results includes progress rate resultson various function classes and convergence proofs for evolution strategies.
Genetic Algorithm II
Genetic Algorithm III
Genetic Algorithm IV
Genetic Algorithm V
Genetic Algorithm VI
 

Download Files

lecture-11-12.pdf (3.02 MB )

homework-week-11-12.doc (0.68 MB )