Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Theoretical Validation of the Latent Optimally Partitioned-$\ell_2/\ell_1$ Penalty with Application to Angular Power Spectrum Estimation

Hiroki Kuroda, Renato Luis Garrido Cavalcante, Masahiro Yukawa

TL;DR

The paper addresses recovering block-sparse signals with unknown block partitions by proving that the latent optimally partitioned (LOP-$\\ell_2/\\ell_1$) penalty selects a partition with blocks of similar magnitudes, yielding a faithful mixed $\\ell_{2}/\\ell_{1}$ norm on the desired partition under a suitable $\\beta_{\\alpha}$ regime. It then applies this insight to angular power spectrum (APS) estimation in MIMO, formulating a linear inverse problem from the sample channel covariance and integrating a dataset of past APS, enhanced by a generalized Moreau envelope to mitigate underestimation. A convex optimization framework with a data fidelity term, a data-driven prior, and the GME-LOP penalty is proposed, with a principled condition on $\\mathbf{B}$ ensuring convexity. Numerical simulations on 3GPP-based scenarios demonstrate significant accuracy gains in APS estimation when exploiting block-sparsity, validating the method's practical impact for moderate-array MIMO systems.

Abstract

This paper demonstrates that, in both theory and practice, the latent optimally partitioned (LOP)-$\ell_2/\ell_1$ penalty is effective for exploiting block-sparsity without knowledge of the concrete block structure. More precisely, we first present a novel theoretical result showing that the optimized block partition in the LOP-$\ell_2/\ell_1$ penalty satisfies a condition required for accurate recovery of block-sparse signals. Motivated by this result, we present a new application of the LOP-$\ell_2/\ell_1$ penalty to estimation of angular power spectrum, which is block-sparse with unknown block partition, in MIMO communication systems. Numerical simulations show that the proposed use of block-sparsity with the LOP-$\ell_2/\ell_1$ penalty significantly improves the estimation accuracy of the angular power spectrum.

Theoretical Validation of the Latent Optimally Partitioned-$\ell_2/\ell_1$ Penalty with Application to Angular Power Spectrum Estimation

TL;DR

The paper addresses recovering block-sparse signals with unknown block partitions by proving that the latent optimally partitioned (LOP-) penalty selects a partition with blocks of similar magnitudes, yielding a faithful mixed norm on the desired partition under a suitable regime. It then applies this insight to angular power spectrum (APS) estimation in MIMO, formulating a linear inverse problem from the sample channel covariance and integrating a dataset of past APS, enhanced by a generalized Moreau envelope to mitigate underestimation. A convex optimization framework with a data fidelity term, a data-driven prior, and the GME-LOP penalty is proposed, with a principled condition on ensuring convexity. Numerical simulations on 3GPP-based scenarios demonstrate significant accuracy gains in APS estimation when exploiting block-sparsity, validating the method's practical impact for moderate-array MIMO systems.

Abstract

This paper demonstrates that, in both theory and practice, the latent optimally partitioned (LOP)- penalty is effective for exploiting block-sparsity without knowledge of the concrete block structure. More precisely, we first present a novel theoretical result showing that the optimized block partition in the LOP- penalty satisfies a condition required for accurate recovery of block-sparse signals. Motivated by this result, we present a new application of the LOP- penalty to estimation of angular power spectrum, which is block-sparse with unknown block partition, in MIMO communication systems. Numerical simulations show that the proposed use of block-sparsity with the LOP- penalty significantly improves the estimation accuracy of the angular power spectrum.

Paper Structure

This paper contains 9 sections, 2 theorems, 14 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Desired block partition for the mixed $\ell_2/\ell_1$ norm
  4. The LOP-$\ell_2/\ell_1$ penalty and its known properties
  5. Theoretical guarantees for the selection of desired blocks with the LOP-$\ell_2/\ell_1$ penalty
  6. Application of the LOP-$\ell_2/\ell_1$ penalty for APS Estimation in MIMO systems
  7. Problem formulation
  8. Proposed method for APS estimation
  9. Simulations

Key Result

Theorem 1

Fix $\bm{x} \in \mathbb{R}^{N}$ and $\alpha \in \mathbb{R}_{++}$ arbitrarily, and set $\beta_{\alpha} \in \mathbb{R}_{+}$ such that the equivalence between eq:ProposedConvexLOPpenalty and eq:LOP_EquivalentAddForm holds. Define $\mathcal{J} := \mathrm{supp}(\bm{x})$, and let $(\mathcal{B}_k^{\ast})_{ where $\bm{1}_{|\mathcal{B}^{\ast}_k|}$ is the all-one vector of size $|\mathcal{B}^{\ast}_k|$. Sup

Figures (1)

  • Figure 1: NMSE versus the number of antennas.

Theorems & Definitions (6)

  • Theorem 1
  • Remark 1
  • Remark 2
  • Remark 3
  • Example 1
  • Proposition 1