A New Adaptive Balanced Augmented Lagrangian Method with Application to ISAC Beamforming Design

Jiageng Wu,Bo Jiang,Xinxin Li,Ya-Feng Liu,Jianhua Yuan

2024-10-20

Abstract:In this paper, we consider a class of convex programming problems with linear equality constraints, which finds broad applications in machine learning and signal processing. We propose a new adaptive balanced augmented Lagrangian (ABAL) method for solving these problems. The proposed ABAL method adaptively selects the stepsize parameter and enjoys a low per-iteration complexity, involving only the computation of a proximal mapping of the objective function and the solution of a linear equation. These features make the proposed method well-suited to large-scale problems. We then custom-apply the ABAL method to solve the ISAC beamforming design problem, which is formulated as a nonlinear semidefinite program in a previous work. This customized application requires careful exploitation of the problem's special structure such as the property that all of its signal-to-interference-and-noise-ratio (SINR) constraints hold with equality at the solution and an efficient computation of the proximal mapping of the objective function. Simulation results demonstrate the efficiency of the proposed ABAL method.

Signal Processing,Information Theory,Optimization and Control

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to solve convex programming problems with linear equality constraints in the fields of machine learning and signal processing. Specifically, the author proposes a new Adaptive Balanced Augmented Lagrangian (ABAL) method to solve these problems. This method is characterized by being able to adaptively select the step - size parameter, and has a low complexity in each iteration, only involving the approximate mapping calculation of the objective function and the solution of linear equations. These features make this method very suitable for large - scale problems.

In addition, the paper also applies the proposed ABAL method to the Intelligent Sensing and Communication (ISAC) beamforming design problem. The ISAC beamforming design problem is formulated as a nonlinear semidefinite programming problem. By taking advantage of the special structure of the problem (for example, all Signal - to - Interference - plus - Noise Ratio (SINR) constraints are satisfied in the form of equality at the solution), and efficiently calculating the approximate mapping of the objective function, the paper demonstrates the efficiency of the proposed method in solving the ISAC beamforming design problem.

In summary, the main contributions of this paper include:
1. Proposing a new Adaptive Balanced Augmented Lagrangian (ABAL) method for solving convex programming problems with linear equality constraints.
2. Applying this method to the ISAC beamforming design problem and demonstrating its efficiency in this application.

The following are the key formulas involved in the paper:

- The mathematical expression of the convex programming problem is:
\[
\min_{u \in \mathbb{C}^n} f(u) \quad \text{s.t.} \quad Du = b,
\]
where \( f: \mathbb{C}^n \to \mathbb{R} \cup \{+\infty\} \) is a proper closed convex function, \( D \in \mathbb{C}^{m \times n} \) and \( b \in \mathbb{C}^m \) are given complex matrices and vectors.

- The definition of the approximate mapping is:
\[
\text{prox}_{\tau f}(v) := \arg\min_{u \in \mathbb{C}^n} \left\{ f(u) + \frac{1}{2\tau} \| u - v \|^2 \right\}.
\]

- The augmented Lagrangian function is:
\[
L_\beta(u; \lambda) = f(u) + \langle Du - b, \lambda \rangle + \frac{\beta}{2} \| Du - b \|^2,
\]
where \( \lambda \in \mathbb{C}^m \) is the Lagrange multiplier related to the equality constraint \( Du = b \), and \( \beta > 0 \) is the penalty parameter.

Through these methods and techniques, the paper successfully solves large - scale convex programming problems and shows significant advantages in ISAC beamforming design.