Diffusion Recursive Minimum Error Entropy Algorithm

Abstract

Distributed estimation algorithms under the mean square error (MSE) criterion have received a lot of attention due to their simplicity and good estimation performance for distributed estimates in Gaussian noise environments. However, when the noise does not obey the Gaussian distribution, the performance of these algorithms can degrade seriously. In this paper, a robust distributed estimation algorithm, called the diffusion recursive minimum error entropy (DRMEE), is proposed by combining the diffusion strategy and the minimum error entropy (MEE) criterion. Since MEE criterion has been proved to be insensitive to many types of non-Gaussian noise models, the proposed algorithm is expected to improve the robustness of distributed estimation algorithms under the MSE criterion, significantly. The superior performance of DRMEE is confirmed by simulation results in the scenario of system identification with two multi-peak distribution noise environments.