Diffusion probabilistic LMS algorithm

TL;DR

This paper introduces the diffusion probabilistic LMS (DPLMS), a new robust estimation algorithm that combines diffusion strategies with probabilistic modeling, outperforming existing algorithms under impulsive interference.

Contribution

The paper presents the DPLMS algorithm, integrating diffusion strategies with probabilistic LMS, providing enhanced robustness and theoretical analysis of stability and complexity.

Findings

DPLMS outperforms existing algorithms in robustness to impulsive interference.

Theoretical analysis confirms stability and manageable computational complexity.

Simulation results demonstrate improved coefficient identification in complex environments.

Abstract

In this paper, a novel diffusion estimation algorithm is proposed from a probabilistic perspective by combining diffusion strategy and the probabilistic least-mean-squares (PLMS) at all agents. The proposed method diffusion probabilistic LMS (DPLMS) is more robust to input signal and impulsive interference than the DSE-LMS, DRVSSLMS and DLLAD algorithms. Instead of minimizing the estimate error, the DPLMS algorithm is derived from approximating the posterior distribution with an isotropic Gaussian distribution. The stability of mean performance and computational complexity are analyzed theoretically. Results from the simulation indicate that the DPLMS algorithm is more robust to input signal and impulsive interference than the DSE-LMS, DRVSSLMS and DLLAD algorithms. These results suggest that the DPLMS algorithm can perform better in identifying the unknown coefficients under the complex and changeable impulsive interference environments.