Abstract:This paper studies a tensor-structured linear regression model with a scalar response variable and tensor-structured predictors, such that the regression parameters form a tensor of order $d$ (i.e., a $d$-fold multiway array) in $\mathbb{R}^{n_1 \times n_2 \times \cdots \times n_d}$. It focuses on the task of estimating the regression tensor from $m$ realizations of the response variable and the predictors where $m\ll n = \prod \nolimits_{i} n_i$. Despite the seeming ill-posedness of this problem, it can still be solved if the parameter tensor belongs to the space of sparse, low Tucker-rank tensors. Accordingly, the estimation procedure is posed as a non-convex optimization program over the space of sparse, low Tucker-rank tensors, and a tensor variant of projected gradient descent is proposed to solve the resulting non-convex problem. In addition, mathematical guarantees are provided that establish the proposed method linearly converges to an appropriate solution under a certain set of conditions. Further, an upper bound on sample complexity of tensor parameter estimation for the model under consideration is characterized for the special case when the individual (scalar) predictors independently draw values from a sub-Gaussian distribution. The sample complexity bound is shown to have a polylogarithmic dependence on $\bar{n} = \max \big\{n_i: i\in \{1,2,\ldots,d \} \big\}$ and, orderwise, it matches the bound one can obtain from a heuristic parameter counting argument. Finally, numerical experiments demonstrate the efficacy of the proposed tensor model and estimation method on a synthetic dataset and a collection of neuroimaging datasets pertaining to attention deficit hyperactivity disorder. Specifically, the proposed method exhibits better sample complexities on both synthetic and real datasets, demonstrating the usefulness of the model and the method in settings where $n \gg m$.

Low-Rank tensor regression: Scalability and applications

Online Robust Low-Rank Tensor Learning.

Online Robust Low-Rank Tensor Modeling for Streaming Data Analysis

Sparse and Low-Rank High-Order Tensor Regression via Parallel Proximal Method

Online Tensor Inference

Support Vector Machine Based On Low-Rank Tensor Train Decomposition For Big Data Applications

Learning from Multiway Data: Simple and Efficient Tensor Regression

Tensor-on-Tensor Regression: Riemannian Optimization, Over-parameterization, Statistical-computational Gap, and Their Interplay

Multilinear Low-Rank Tensors on Graphs & Applications

Greedy Learning of Generalized Low-Rank Models.

Low-Rank Tensor Learning by Generalized Nonconvex Regularization

Learning of Generalized Low-Rank Models: A Greedy Approach.

ISLET: Fast and Optimal Low-rank Tensor Regression via Importance Sketching

A literature survey of low-rank tensor approximation techniques

A Solution for Large Scale Nonlinear Regression with High Rank and Degree at Constant Memory Complexity via Latent Tensor Reconstruction

Unsupervised learning low-rank tensor from incomplete and grossly corrupted data

Low-rank Tensor Ring Learning for Multi-Linear Regression

Low-Rank Modeling and Its Applications in Image Analysis

Tensor Regression Using Low-rank and Sparse Tucker Decompositions

Low-Rank Tensor Train Coefficient Array Estimation for Tensor-on-Tensor Regression

Low Rank Metric Learning via Accelerated Gradient Optimization