TY - GEN
T1 - Q-Learning-Based Adaptive Bacterial Foraging Optimization
AU - Niu, Ben
AU - Xue, Bowen
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020/11/11
Y1 - 2020/11/11
N2 - As a common biological heuristic algorithm, Bacterial Foraging Optimization (BFO) is often used to solve optimization problems. Aiming at increasing the solution accuracy and convergence performance while enhancing the capability of individual’s self-learning and exploration, a Q-Learning-Based Adaptive Bacterial Foraging Optimization (QABFO) is proposed in this paper. The basic chemotaxis, reproduction and elimination/dispersal operations in standard BFO are redesigned in a Q-learning mechanism, and the Q-table will be updated on the basis of the changed fitness values in each iteration. In chemotaxis operation, modified search direction and secondary cruising mechanism are introduced into tumbling and swimming behaviors, with the purpose of improving the search efficiency and balancing local and global search. Additionally, a generation skipping adaptive reproduction is designed to control the accuracy and convergence of QABFO. Experimental results demonstrate that compared with BFO, PSO and GA, the proposed algorithm performs better in terms of accuracy and stability on most of the test functions and can effectively improve the premature convergence problem due to the original reproduction operation in BFO.
AB - As a common biological heuristic algorithm, Bacterial Foraging Optimization (BFO) is often used to solve optimization problems. Aiming at increasing the solution accuracy and convergence performance while enhancing the capability of individual’s self-learning and exploration, a Q-Learning-Based Adaptive Bacterial Foraging Optimization (QABFO) is proposed in this paper. The basic chemotaxis, reproduction and elimination/dispersal operations in standard BFO are redesigned in a Q-learning mechanism, and the Q-table will be updated on the basis of the changed fitness values in each iteration. In chemotaxis operation, modified search direction and secondary cruising mechanism are introduced into tumbling and swimming behaviors, with the purpose of improving the search efficiency and balancing local and global search. Additionally, a generation skipping adaptive reproduction is designed to control the accuracy and convergence of QABFO. Experimental results demonstrate that compared with BFO, PSO and GA, the proposed algorithm performs better in terms of accuracy and stability on most of the test functions and can effectively improve the premature convergence problem due to the original reproduction operation in BFO.
KW - bacterial foraging optimization
KW - generation skipping adaptive reproduction
KW - Q-learning
UR - http://www.scopus.com/inward/record.url?scp=85097189146&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-62460-6_29
DO - 10.1007/978-3-030-62460-6_29
M3 - Conference contribution
AN - SCOPUS:85097189146
SN - 9783030624590
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 327
EP - 337
BT - Machine Learning for Cyber Security
A2 - Chen, Xiaofeng
A2 - Yan, Hongyang
A2 - Yan, Qiben
A2 - Zhang, Xiangliang
PB - Springer Cham
CY - Cham
T2 - 3rd International Conference on Machine Learning for Cyber Security, ML4CS 2020
Y2 - 8 October 2020 through 10 October 2020
ER -