Analysis and Design of Irregular Graphs for Node-Based Verification-Based Recovery Algorithms in Compressed Sensing

TL;DR

This paper introduces a probabilistic analysis method for node-based verification algorithms in compressed sensing, enabling the design of irregular graphs that significantly improve recovery performance over regular graphs.

Contribution

It proposes a simpler, more accurate asymptotic analysis technique for irregular graphs in compressed sensing, facilitating the design of more effective sensing graphs.

Findings

Irregular graphs outperform regular graphs in signal recovery.

The analysis accurately predicts performance for large signal dimensions.

Designed irregular graphs recover up to 40% more non-zero elements.

Abstract

In this paper, we present a probabilistic analysis of iterative node-based verification-based (NB-VB) recovery algorithms over irregular graphs in the context of compressed sensing. Verification-based algorithms are particularly interesting due to their low complexity (linear in the signal dimension $n$). The analysis predicts the average fraction of unverified signal elements at each iteration $\ell$ where the average is taken over the ensembles of input signals and sensing matrices. The analysis is asymptotic ($n \rightarrow \infty$) and is similar in nature to the well-known density evolution technique commonly used to analyze iterative decoding algorithms. Compared to the existing technique for the analysis of NB-VB algorithms, which is based on numerically solving a large system of coupled differential equations, the proposed method is much simpler and more accurate. This allows us to design irregular sensing graphs for such recovery algorithms. The designed irregular graphs outperform the corresponding regular graphs substantially. For example, for the same recovery complexity per iteration, we design irregular graphs that can recover up to about 40% more non-zero signal elements compared to the regular graphs. Simulation results are also provided which demonstrate that the proposed asymptotic analysis matches the performance of recovery algorithms for large but finite values of $n$.