Ensemble-based control of evolutionary optimization algorithms

This paper presents a study on how the intrinsic search parameters of an evolutionary optimization algorithm can be automatically controlled. It will be shown that only a small search parameter window ensures good optimization results. This evolutionary window, enclosing effective values for the mutation rate and temperature, can be adapted to by carefully steering the ensemble’s fitness dispersion. A control sensor based on an entropy measure is introduced to achieve this goal. The efficiency of the proposed control method will be tested by optimizing artificial sequences such as the well-known low autocorrelated binary strings and natural sequences including RNA.