Nonsubsampled Graph Filter Banks and Distributed Implementation

TL;DR

This paper introduces nonsubsampled graph filter banks (NSGFBs) for distributed graph data processing, providing perfect reconstruction, noise control, and efficient implementation on large graphs.

Contribution

It proposes algebraic and optimization methods for constructing synthesis filter banks ensuring perfect reconstruction and noise control in NSGFBs, with a distributed implementation approach.

Findings

NSGFBs achieve perfect signal reconstruction in noiseless settings.

Proposed methods control resonance effects under bounded noise.

Distributed denoising technique effectively suppresses noise.

Abstract

In this paper, we consider nonsubsampled graph filter banks (NSGFBs) to process data on a graph in a distributed manner. Given an analysis filter bank with small bandwidth, we propose algebraic and optimization methods of constructing synthesis filter banks such that the corresponding NSGFBs provide a perfect signal reconstruction in the noiseless setting. Moreover, we prove that the proposed NSGFBs can control the resonance effect in the presence of bounded noise and they can limit the influence of shot noise primarily to a small neighborhood of its location on the graph. For an NSGFB on a graph of large size, a distributed implementation has a significant advantage, since data processing and communication demands for the agent at each vertex depend mainly on its neighboring topology. In this paper, we propose an iterative distributed algorithm to implement the proposed NSGFBs. Based on NSGFBs, we also develop a distributed denoising technique which is demonstrated to have satisfactory performance on noise suppression.