A Variable Step-Size MSD-Based Diffusion LMS Algorithm With Noisy Input and Quantization Link Noise

This paper presents a variable step-size diffusion least-mean-square algorithm that considers both input and quantization noise in digital communication, particularly in wireless sensor networks. Whereas link noise has simply been modeled as additive white Gaussian noise in the previous algorithms, this study considers uniform quantization effects during weight vector transmission. The proposed combination protocol utilizes mean-square-deviation values to set an upper bound, dynamically adjusting the combination coefficients. As a result, it ensures energy-efficient operation even in low-bit allocation environments. Experimental results indicate that the proposed algorithm surpasses the existing algorithms and maintains performance in changing environments through reset algorithms.