Probabilistic Optimization Algorithms for numerical function optimization problems

This paper proposes a novel optimization algorithm called cellular probabilistic optimization algorithms (CPOA) based on the probabilistic representation of solutions for real coded problems. In place of binary integers, the basic unit of information here is a probability density function. This probabilistic coding allows superposition of states for a more efficient algorithm. This probabilistic representation enables the algorithm to climb the hills in the search space. Furthermore, the cellular structure of the proposed algorithm aims to provide an appropriate tradeoff between exploitation and exploration. The proposed algorithm is tested on several numeric benchmark function optimization problems. Experimental results show that the performance of CPEA is improved when compared with other evolutionary algorithms like particle swarm optimization (PSO) and genetic algorithms (GA). Furthermore, this improvement becomes particularly more significant for problems with higher dimension.