Graph Signal Estimation Based on Maximum Correntropy Criterion (The Work of A. Chandrasekar Was Supported by the MATRICS, SERB, India, Under Grant MTR/2021/000405)

In order to estimate the graph signal from the small part of samples, a novel adaptive filtering method based on maximum correntropy criterion is proposed in this study. The proposed approach is resistant to environments with impulsive noise. The performance enhancement of the proposed technique is well demonstrated by the simulation results carried out in the context of weather data.

I. Introduction

Graphs are found to be effective in representation of all data types which results in the popularity of graph based approach for several applications. Recent developments in graph signal processing have paved the way for solving the problem of processing graph data. It is well known that adaptive filtering algorithms have proven to be effective in estimation and tracking of signals that vary with time. Incorporating the adaptive filtering approach into graph signal processing, results in adaptive Graph Signal Processing (GSP) algorithms for effective online estimation of signals represented in graph.