Shrinking Three Stage Optimal Memetic Exploration

The Ockham’s razor in Memetic Computing states that optimization algorithms composed of few, simple, and tailored components can be very efficient, if properly designed. If the designer is aware of the role and effect of each algorithmic component an high performance can be easily obtained. Following this principle, this paper proposes a novel algorithm for numerical optimization. The proposed algorithm, namely Shrinking Three Stage Optimal Memetic Exploration (S-3SOME), performs the progressive perturbation of a candidate solution by alternating three search operators, the first is a stochastic global search, the second is random sampling within progressive narrowing hypervolume, the third is a deterministic local search. The proposed S-3SOME is an efficient scheme which outperforms, for the considered problems, a similar scheme proposed in literature and, despite its simplicity, is competitive with complex population-based algorithms which require massive overhead and memory employment