TY - GEN
T1 - Co-operative Prediction Strategy for Solving Dynamic Multi-Objective Optimization Problems
AU - Zhao, Zhihao
AU - Gu, Fangqing
AU - Cheung, Yiu Ming
N1 - Funding Information:
This work was supported by the National Science Foundation of China under Grants: 61672444 and 61703108, by Hong Kong Baptist University (HKBU), Research Committee, Initiation Grant, Faculty Niche Research Areas (IG-FNRA) 2018/19 under Grant RC-FNRA-IG/18-19/SCI/03, by the Innovation and Technology Fund of Innovation and Technology Commission of the Government of the Hong Kong SAR under Project ITS/339/18, by the SZSTI under Grant JCYJ20160531194006833, and by the Natural Science Foundation of Guangdong Province 2017A030310467.
PY - 2020/7
Y1 - 2020/7
N2 - Prediction-based evolutionary multi-objective optimization algorithm is one of the most popular optimization algorithms for solving dynamic multi-objective optimization problem. It uses time-series models to predict the future Pareto set based on the past solutions. However, the dimension of the decision variables may be too high to predict. Moreover, a relatively small variance in decision variables may lead to a large difference in the objective space. The optimized Pareto front (PF) may be far from the desired output. To solve these problems, this paper proposes a new co-operative prediction method, which predicts not only the Pareto solution (PS), but also a hyper-plane as an approximation of the prediction of the PF in the objective space. The hyper-plane is used to guide the search process and accelerate the convergence. We compare the proposed algorithm with three existing dynamic optimization algorithms. Experimental results show the effectiveness of the proposed algorithm.
AB - Prediction-based evolutionary multi-objective optimization algorithm is one of the most popular optimization algorithms for solving dynamic multi-objective optimization problem. It uses time-series models to predict the future Pareto set based on the past solutions. However, the dimension of the decision variables may be too high to predict. Moreover, a relatively small variance in decision variables may lead to a large difference in the objective space. The optimized Pareto front (PF) may be far from the desired output. To solve these problems, this paper proposes a new co-operative prediction method, which predicts not only the Pareto solution (PS), but also a hyper-plane as an approximation of the prediction of the PF in the objective space. The hyper-plane is used to guide the search process and accelerate the convergence. We compare the proposed algorithm with three existing dynamic optimization algorithms. Experimental results show the effectiveness of the proposed algorithm.
KW - Dynamic Multiobjective Optimization
KW - Evolutionary Algorithms
UR - http://www.scopus.com/inward/record.url?scp=85092028235&partnerID=8YFLogxK
U2 - 10.1109/CEC48606.2020.9185721
DO - 10.1109/CEC48606.2020.9185721
M3 - Conference proceeding
AN - SCOPUS:85092028235
T3 - 2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings
BT - 2020 IEEE Congress on Evolutionary Computation, CEC 2020 - Conference Proceedings
PB - IEEE
T2 - 2020 IEEE Congress on Evolutionary Computation, CEC 2020
Y2 - 19 July 2020 through 24 July 2020
ER -