Simple probabilistic population based optimization for combinatorial optimization
Ying-Chi Lin, Martin Middendorf · 2013
A new scheme is proposed for the design of probabilistic population based optimization algorithms for solving combinatorial optimization problems. The new scheme, Simple Probabilistic Population Based Optimization scheme (SPPBO), is used also to classify existing metaheuristics, e.g., the Population-based Ant Colony Optimization algorithm (PACO) and the Simplified Swarm Optimization algorithm (SSO). The classification shows the close relationship between PACO and SSO. This fact has not been recognized in the literature so far. SPPBO is also used to identify new metaheuristics that come up naturally as variants and combinations of PACO and SSO. An experimental study is done to evaluate and compare the different algorithms when applied to the Traveling Salesperson Problem. The results show which parts of the algorithms are helpful for obtaining a good optimization behaviour. In addition to the original PACO and SSO algorithms also some of the new combinations perform very well.