Second Order Estimation of Distribution Algorithms Based on Kalman Filter
Zhong Wei · Chinese Journal of Computers · 2004
Estimation of Distribution Algorithms (EDAs) are new evolutionary algorithms based on probabilistic model and have become a new focus in the field of evolutionary computation. From the view point of Kalman filter, EDA actually is a filter with single sensor, so its stability is poor and it is prone to be trapped in the local optima of the objective functions. To overcome these disadvantages, authors enhance its performance with Kalman filtering technique and propose a new algorithm, second order estimation of distribution algorithm based on Kalman filter. In this method, population is divided into several sub-populations, and a second order EDA for each sub-population is used to estimate the information of its state. Then, a Kalman filter is used to fuse the information so that more accurate state can be obtained. Finally, the information fused is fed back to each sub-population. Experimental results demonstrate that the algorithm outperforms available second order algorithm greatly both in the stability and the global search ability.