Phase Transitions with Structured Sparsity

TL;DR

This paper investigates the phase transition phenomena in structured sparse signals, such as block and tree structures, and derives thresholds for their recovery in high-dimensional convex optimization contexts.

Contribution

It extends phase transition analysis to structured sparse signals, providing threshold expressions for block and tree structures in compressed sensing.

Findings

Derived strong threshold expressions for block-structured sparse signals.

Analyzed phase transition thresholds for tree-structured sparse signals.

Enhanced understanding of structured sparsity effects on signal recovery thresholds.

Abstract

In the field of signal processing, phase transition phenomena have recently attracted great attention. Donoho's work established the signal recovery threshold using indicators such as restricted isotropy (RIP) and incoherence and proved that phase transition phenomena occur in compressed sampling. Nevertheless, the phase transition phenomenon of structured sparse signals remains unclear, and these studies mainly focused on simple sparse signals. Signals with a specific structure, such as the block or tree structures common in real-world applications, are called structured sparse signals. The objectives of this article are to study the phase transition phenomenon of structured sparse signals and to investigate how structured sparse signals affect the phase transition threshold. It begins with a summary of the common subspace of structured sparse signals and the theory of high-dimensional convex polytope random projections. Next, the strong threshold expression of block-structured and tree-structured sparse signals is derived after examining the weak and strong thresholds of structured sparse signals.