Abstract:Noiseless compressive sensing is a protocol that enables undersampling and later recovery of a signal without loss of information. This compression is possible because the signal is usually sufficiently sparse in a given basis. Currently, the algorithm offering the best tradeoff between compression rate, robustness, and speed for compressive sensing is the LASSO (l1-norm bias) algorithm. However, many studies have pointed out the possibility that the implementation of lp-norms biases, with p smaller than one, could give better performance while sacrificing convexity. In this work, we focus specifically on the extreme case of the l0-based reconstruction, a task that is complicated by the discontinuity of the loss. In the first part of the paper, we describe via statistical physics methods, and in particular the replica method, how the solutions to this optimization problem are arranged in a clustered structure. We observe two distinct regimes: one at low compression rate where the signal can be recovered exactly, and one at high compression rate where the signal cannot be recovered accurately. In the second part, we present two message-passing algorithms based on our first results for the l0-norm optimization problem. The proposed algorithms are able to recover the signal at compression rates higher than the ones achieved by LASSO while being computationally efficient.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to explore the Compressed Sensing (CS) problem based on the ℓ₀ -norm. Specifically, it attempts to solve the following problems: 1. **Optimizing Sparse Signal Reconstruction**: - Compressed Sensing is a technique for signal compression by reducing the number of measurements, but it is necessary to ensure lossless information during the recovery process. Usually, the signal is sparse under a certain basis (i.e., most elements are zero), which makes compression possible. - The currently most commonly used algorithm is LASSO (based on the ℓ₁ -norm), but in some cases, using a smaller p -value (such as the ℓ₀ -norm) may provide better performance, although this sacrifices convexity. 2. **Application of Statistical Physics Methods**: - The paper uses the replica method in statistical physics to analyze the solution structure of the ℓ₀ -norm optimization problem. This method can reveal how the solution space is organized and help understand the signal recovery performance under different compression ratios. - Research shows that there are two different regions: the signal can be accurately recovered at a low compression ratio, while it cannot be accurately recovered at a high compression ratio. 3. **Proposing New Algorithms**: - Based on the above - mentioned theoretical analysis, the paper proposes two message - passing algorithms (ASP and ASPo) based on message passing. These algorithms can effectively recover signals at a higher compression ratio and are computationally efficient. - These new algorithms are superior to the existing LASSO algorithm under certain conditions, especially when dealing with high - dimensional sparse signals. 4. **Exploring the Limits of ℓ₀ -norm Optimization**: - The paper also studies the theoretical limits of ℓ₀ -norm optimization and shows the recovery thresholds under different compression ratios and signal sparsities through phase diagrams. - The results show that the recovery performance based on the ℓ₀ -norm is close to the Bayesian Optimal (BO) recovery threshold, which is very surprising because BO requires a complete understanding of the statistical characteristics of the signal generation process, while the ℓ₀ -norm method is relatively more general. In summary, this paper mainly solves the problem of how to use the ℓ₀ -norm for signal recovery in Compressed Sensing, and improves the performance and efficiency of signal recovery through statistical physics methods and the design of new algorithms.

Compressed sensing with l0-norm: statistical physics analysis and algorithms for signal recovery

Weighted LASSO for Sparse Recovery with Statistical Prior Support Information

Stable recovery of sparse signals with coherent tight frames via lp-analysis approach

The phase diagram of compressed sensing with $\ell_0$-norm regularization

Noisy Signal Recovery via Iterative Reweighted L1-Minimization

A Fast Sparse Recovery Algorithm for Compressed Sensing Using Approximate l0 Norm and Modified Newton Method

Robust Sparse Recovery in Impulsive Noise via lp-l1 Optimization

Iteratively Reweighted Algorithms for Compressive Sensing

A Fast Sparse Recovery Algorithm for Compressed Sensing Using Approximate <i>l</i><sub>0</sub> Norm and Modified Newton Method

Robust Sparse Recovery for Compressive Sensing in Impulsive Noise Using ℓp-Norm Model Fitting.

Image Compressive Sensing Recovery Using Adaptively Learned Sparsifying Basis Via L0 Minimization

Statistical physics-based reconstruction in compressed sensing

Recovery of Correlated Row Sparse Signals Using Smoothed L 0 -Norm Algorithm

A Sparse Recovery Algorithm Using the Neurodynamic System with Predefined Time Convergence

A Null-Space-Based Weighted L(1) Minimization Approach to Compressed Sensing

A Fast Sparse Recovery Algorithm for Compressed Sensing Using Approximate ₀ Norm and Modified Newton Method.

Subspace Pursuit for Compressive Sensing Signal Reconstruction

Efficient Tuning-Free $\ell_1$-Regression of Nonnegative Compressible Signals

Globally Sparse and Locally Dense Signal Recovery for Compressed Sensing

A Barzilai-Borwein $l_1$-Regularized Least Squares Algorithm for Compressed Sensing

Sparse signal reconstruction based on iterative smoothed l 0 norm minimization