A noisy multi-objective optimization algorithm based on mean and Wiener filters

摘要

Recently, evolutionary algorithms have made great achievements in multi-objective optimization problems (MOPs), but there is a little research on how to deal with noisy multi-objective optimization problems (NMOPs), which are quite common in real life. The work in this paper attempts to find the commonality of noises in images/signals and NMOPs and analyzes the effects of several classical smoothing techniques in images/signals on NMOPs. A novel denoising algorithm that embeds the mean and Wiener filters into existing multi-objective optimization algorithms is proposed. In the proposed method, resampling is employed to maintain the accuracy of non-dominated solutions and filters are utilized to denoise dominated solutions, where the mean and Wiener filters are conducive to balance convergence and diversity of the population, respectively. This is the first use of filters in NMOPs, and it will promote the application of denoising methods in images/signals to NMOPs. The empirical results show the superiority of filters to denoise problems with low/medium-intensity noises and continuous Pareto front, compared with several state-of-the-art algorithms.