Gaussian Process Machine Learning Based ITO Algorithm
Chuang Ma, Yongjian Yang, Zhanwei Du, Chijun Zhang · 2014
Taking the Gaussian process (GP) regression model as ITO's fluctuation operator, we propose a new mixed algorithm called GITO in order to overcome the local minima problem. Through learning the particles' mobility models, ITO's capacity of local searching and global searching is strengthened. Meanwhile, we give the proof procedure to verify ITO's fluctuation operator and GP are logically equivalent. Finally, the experiments show GITO's better convergence rate and performance.