Self Adaptation of Operator Rates in Evolutionary Algorithms

This work introduces a new evolutionary algorithm that adapts the operator probabilities (rates) while evolves the solution of the problem. Each individual encodes its genetic rates. In every generation, each individual is modified by only one operator that is selected according to the encoded rates. Such rates are updated according to the performance achieved by the offspring (compared to its parent) and a random learning rate. The proposed approach is augmented with a simple transposition operator and tested on a number of benchmark functions.