COMBEO: An evolutionary global optimization framework based on change of measures
Saikat Sarkar, Debasish K. Roy · arXiv (Cornell University) · 2014
An evolutionary global optimization paradigm, code-named COMBEO (Change Of Measure Based Evolutionary Optimization) is proposed. The main thrust in the development is a set of derivative-free directional terms derivable through a change of measures to impose any stipulated conditions aimed at driving the realized design variables (particles) to the global optimum. The directional terms, additively applied to correct the current particles and interpretable as the stochastic equivalent of the Gateaux derivatives conventionally employed in a deterministic search for a local extremum, are computable from the sample statistical moments within a Monte Carlo setup. The generalized setting offered by the new approach also enables one to borrow several basic ideas, used with other global search methods such as the particle swarm or the differential evolution, to be rationally incorporated via a change of measures. The global search may be further aided by imparting to the directional update terms additional layers of random perturbations such as 'scrambling' and 'selection'. Depending on the precise choice of the optimality conditions and the extent of random perturbation, the search can be made either greedy or more exploratory. As numerically demonstrated, the new proposal apparently provides for a more rational, more accurate and faster alternative to most available evolutionary schemes, prominent amongst whom are the differential evolution and the particle swarm optimization.