An efficient EM algorithm for both element-wise and structural missingness in matrix-variate normal mixture models
Abstract
Matrix-variate data with missing entries arise frequently in applications where observations are naturally organized as two-dimensional arrays. Although the matrix normal distribution provides a parsimonious model through its Kronecker covariance structure, standard EM estimation can be computationally expensive because arbitrary missingness patterns typically destroy this separability in the E-step. In this paper, we propose an efficient partial EM algorithm for matrix-variate normal data with missing entries. The proposed method updates the conditional mean and covariance of the missing component through coordinate-wise approximations, avoiding repeated inversion of pattern-specific covariance matrices and avoiding construction of the full vectorized covariance matrix. We further develop a specialized update for submatrix missingness, where the missing-block precision retains a Kronecker product structure, and the covariance update can be carried out independently in the row and column directions. Simulation studies show that the proposed methods substantially reduce computation time compared with exact EM while preserving nearly identical observed-data likelihood across a range of dimensions and missing proportions. A real-data application to hyperspectral image patches demonstrates that the proposed imputation strategy can be embedded within a matrix-variate mixture model for simultaneous imputation and clustering.
Related
- An energy-decreasing algorithm for the finite element approximation of ferronematic equilibrium states and its numerical realization
- Algorithm XXXX: Computation of finite element degree-of-freedom transformation matrices
- An Efficient Vertex Retrieval Algorithm for Ball Polyhedra
- An Efficient iLQR Algorithm for Optimal Control with Bounded States and Inputs: Application to Spacecraft Maneuvers
- ITHUIM: An efficient algorithm for mining inter-transaction high utility itemsets
BibTeX
@misc{lu2026efficient,
title = {{An efficient EM algorithm for both element-wise and structural missingness in matrix-variate normal mixture models}},
author = {Hanzhang Lu and Jeffrey L. Andrews and Ryan P. Browne},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.00616},
eprint = {2609.00616},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.00616},
}
RIS
TY - UNPB
TI - An efficient EM algorithm for both element-wise and structural missingness in matrix-variate normal mixture models
AU - Hanzhang Lu
AU - Jeffrey L. Andrews
AU - Ryan P. Browne
PY - 2026
DA - 2026/09/01
PB - arXiv
DO - 10.48550/arxiv.2609.00616
UR - https://arxiv.org/abs/2609.00616
AB - Matrix-variate data with missing entries arise frequently in applications where observations are naturally organized as two-dimensional arrays. Although the matrix normal distribution provides a parsimonious model through its Kronecker covariance structure, standard EM estimation can be computationally expensive because arbitrary missingness patterns typically destroy this separability in the E-step. In this paper, we propose an efficient partial EM algorithm for matrix-variate normal data with missing entries. The proposed method updates the conditional mean and covariance of the missing component through coordinate-wise approximations, avoiding repeated inversion of pattern-specific covariance matrices and avoiding construction of the full vectorized covariance matrix. We further develop a specialized update for submatrix missingness, where the missing-block precision retains a Kronecker product structure, and the covariance update can be carried out independently in the row and column directions. Simulation studies show that the proposed methods substantially reduce computation time compared with exact EM while preserving nearly identical observed-data likelihood across a range of dimensions and missing proportions. A real-data application to hyperspectral image patches demonstrates that the proposed imputation strategy can be embedded within a matrix-variate mixture model for simultaneous imputation and clustering.
ER -