Low-rank Latent Matrix-factor Prediction Modeling for Generalized High-dimensional Matrix-variate Regression

TL;DR

This paper introduces LaGMaR, a low-rank matrix-factor regression model that efficiently predicts responses from high-dimensional matrix covariates, especially useful for medical imaging data like CT scans.

Contribution

The paper proposes a novel low-rank latent matrix-factor regression model that reduces dimensionality respecting the matrix structure, improving prediction and computational efficiency.

Findings

LaGMaR outperforms existing penalized methods in simulations.

The method accurately predicts COVID-19 status from CT scan biomarkers.

The approach maintains structural information of matrix covariates.

Abstract

Motivated by diagnosing the COVID-19 disease using 2D image biomarkers from computed tomography (CT) scans, we propose a novel latent matrix-factor regression model to predict responses that may come from an exponential distribution family, where covariates include high-dimensional matrix-variate biomarkers. A latent generalized matrix regression (LaGMaR) is formulated, where the latent predictor is a low-dimensional matrix factor score extracted from the low-rank signal of the matrix variate through a cutting-edge matrix factor model. Unlike the general spirit of penalizing vectorization plus the necessity of tuning parameters in the literature, instead, our prediction modeling in LaGMaR conducts dimension reduction that respects the geometry characteristic of intrinsic two-dimensional structure of the matrix covariate and thus avoids iteration. This greatly relieves the computation burden, and meanwhile maintains structural information so that the latent matrix factor feature can perfectly replace the intractable matrix-variate owing to high-dimensionality. The estimation procedure of LaGMaR is subtly derived by transforming the bilinear form matrix factor model onto a high-dimensional vector factor model, so that the method of principle components can be applied. We establish bilinear-form consistency of the estimated matrix coefficient of the latent predictor and consistency of prediction. The proposed approach can be implemented conveniently. Through simulation experiments, prediction capability of LaGMaR is shown to outperform existing penalized methods under diverse scenarios of generalized matrix regressions. Through the application to a real COVID-19 dataset, the proposed approach is shown to predict efficiently the COVID-19.