A Multi-population-Based Algorithm with Different Ways of Subpopulations Cooperation.

Metaheuristic methods are designed to solve continuous and discrete problems. Such methods include population based algorithms (PBAs). They are distinguished by the flexibility of defining the fitness function, therefore they are a good alternative to gradient methods. However, creating new variants of PBAs that work similarly and differ in detail might be problematic. Therefore, it is interesting to combine existing PBAs in order to increase their effectiveness. One of the hybrid methods is the Multi-population Nature-Inspired Algorithm (MNIA), which uses search operators from different PBAs. The formula of MNIA’s operation is based on the appropriate cooperation of its subpopulations. That is why in this paper we focus on expanding MNIAs with various schemes of such cooperation. In particular, we analyze various combinations of migration models, intervals, and topologies. The proposed solutions were tested and compared using generally known benchmark functions. The obtained results showed an advantage of certain patterns of cooperation of the - subpopulations, which confirmed the validity of the adopted assumptions.