Bivariate Matrix-valued Linear Regression (BMLR): Finite-sample performance under Identifiability and Sparsity Assumptions

Neural Information Processing Systems 

This paper studies a bilinear matrix-valued regression model where both predictors and responses are matrices.