Abstract:Affine sum-of-ranks minimization (ASRM) generalizes the affine rank minimization (ARM) problem from matrices to tensors. Here, the interest lies in the ranks of a family $\mathcal{K}$ of different matricizations. Transferring our priorly discussed results on asymptotic log-det rank minimization, we show that iteratively reweighted least squares with weight strength $p = 0$ remains a, theoretically and practically, particularly viable method denoted as $\mathrm{IRLS}$-$0\mathcal{K}$. As in the matrix case, we prove global convergence of asymptotic minimizers of the log-det sum-of-ranks function to desired solutions. Further, we show local convergence of $\mathrm{IRLS}$-$0\mathcal{K}$ in dependence of the rate of decline of the therein appearing regularization parameter $\gamma \searrow 0$. For hierarchical families $\mathcal{K}$, we show how an alternating version ($\mathrm{AIRLS}$-$0\mathcal{K}$, related to prior work under the name $\mathrm{SALSA}$) can be evaluated solely through tensor tree network based operations. The method can thereby be applied to high dimensions through the avoidance of exponential computational complexity. Further, the otherwise crucial rank adaption process becomes essentially superfluous even for completion problems. In numerical experiments, we show that the therefor required subspace restrictions and relaxation of the affine constraint cause only a marginal loss of approximation quality. On the other hand, we demonstrate that $\mathrm{IRLS}$-$0\mathcal{K}$ allows to observe the theoretical phase transition also for generic tensor recoverability in practice. Concludingly, we apply $\mathrm{AIRLS}$-$0\mathcal{K}$ to larger scale problems.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the Affine Sum - of - Ranks Minimization (ASRM) problem of tensors. Specifically, ASRM is the generalization of the Affine Rank Minimization (ARM) problem of matrices to tensors. Its core objective is to minimize the sum of the ranks of a set of specific matrixizations under linear constraint conditions. #### Problem background and motivation - **Affine Rank Minimization (ARM)**: In the case of matrices, the ARM problem refers to finding a matrix with as low a rank as possible under given linear measurements. - **Tensor Extension**: ASRM generalizes this problem to tensors, considering the problem of minimizing the sum of the ranks of different matrixizations. Here, matrixization refers to rearranging a tensor into a matrix form, thereby using the rank of the matrix to measure the complexity of the tensor. #### Specific problem description Given: - A $d$-dimensional tensor $X\in\mathbb{R}^{n_1\times n_2\times\cdots\times n_d}$ - A linear operator $L:\mathbb{R}^{n_1\times n_2\times\cdots\times n_d}\to\mathbb{R}^{\ell}$ - A measurement value $y\in\text{image}(L)$ The goal is to solve the following optimization problem: \[ \argmin_{X\in\mathbb{R}^{n_1\times n_2\times\cdots\times n_d}}\sum_{J\in K}\text{rank}(X[J])\quad\text{subject to}\quad L(X) = y \] where $K\subseteq\{J\subset[d]\mid J\neq\emptyset\}$, and $X[J]$ represents a matrixized form of the tensor $X$. #### Solutions The authors introduce the asymptotic log - det minimization method, combined with the Iteratively Reweighted Least Squares (IRLS), and propose an effective algorithm framework. Specifically: - **Global convergence**: Prove the convergence of the global optimal solution. - **Local convergence**: Analyze the local convergence behavior when the regularization parameter $\gamma\to0$. - **Alternating version**: Propose the Alternating IRLS (AIRLS) suitable for high - dimensional problems, and show its application in tree - like tensor networks. #### Main contributions - Generalize the ARM problem of matrices to the ASRM problem of tensors. - Propose the asymptotic log - det minimization method and prove its theoretical and practical feasibility. - Demonstrate the superior performance of the AIRLS method in high - dimensional tensor recovery problems. Through these works, the paper provides a complete theoretical basis and computational method for the tensor rank minimization problem.

Asymptotic Log-Det Sum-of-Ranks Minimization via Tensor (Alternating) Iteratively Reweighted Least Squares

CONVERGENCE AND STABILITY OF ITERATIVELY REWEIGHTED LEAST SQUARES FOR LOW-RANK MATRIXRECOVERY

Smoothed Low Rank and Sparse Matrix Recovery by Iteratively Reweighted Least Squares Minimization

On the Global Convergence of the Alternating Least Squares Method for Rank-One Approximation to Generic Tensors

Alternating Iteratively Reweighted Minimization Algorithms for Low-Rank Matrix Factorization

Conjugate gradient acceleration of iteratively re-weighted least squares methods

Iteratively re-weighted least squares minimization for sparse recovery

Log-Sum Regularized Kaczmarz Algorithms for High-Order Tensor Recovery

Convergence and stability of iteratively reweighted least squares for low-rank matrix recovery

A New Nonconvex Strategy to Affine Matrix Rank Minimization Problem

Efficient Alternating Least Squares Algorithms for Low Multilinear Rank Approximation of Tensors

Structured Low-Rank Tensors for Generalized Linear Models

LatLRR for subspace clustering via reweighted Frobenius norm minimization

Low-tubal-rank Tensor Completion using Alternating Minimization

An Iterative Reweighted Method for Tucker Decomposition of Incomplete Multiway Tensors

Structure-preserving low multilinear rank approximation of antisymmetric tensors

A Riemannian rank-adaptive method for higher-order tensor completion in the tensor-train format

Precise asymptotics of reweighted least-squares algorithms for linear diagonal networks

CP decomposition and low-rank approximation of antisymmetric tensors

Efficient Alternating Minimization with Applications to Weighted Low Rank Approximation

Learning-Based Low-Rank Approximations