Bacterial Foraging Algorithm Based on Reinforcement Learning for Continuous Optimizations

Abstract

Inspired by the adaptive learning and foraging behaviors of the bacteria colony, this paper proposes a re-structured bacterial colony foraging optimizer (RBCFO) based on reinforcement learning (RL) and self-adaptive search strategy for continuous optimizations. The algorithm aims to enhance the individual search efficiency during the evolution process via exploiting the RL mechanism in the multi-operations decision level and the adaptive search strategy in the single-operation level. Specifically, in the multi-operations decision level, the operation of each bacterium is determined by RL in an optimal manner. In the single-operation level (i.e., chemotaxis), each bacterium adaptively varies its own run-length unit and exchange information (i.e., cell-to-cell crossover) to dynamically balance exploration and exploitation during the search process. Then the proposed algorithm is evaluated against several state-of-the-art algorithms on a set of continuous benchmark instances. Experimental results verify the significant superiority of the proposed algorithm.