Chain-linked multiple matrix integration via embedding alignment

TL;DR

This paper introduces CMMI, a novel method for integrating multiple noisy submatrices with block-wise missing data to accurately reconstruct the full matrix, using embedding alignment and aggregation techniques.

Contribution

The paper proposes the Chain-linked Multiple Matrix Integration (CMMI) method, a new approach for matrix completion from block-wise missing data using embedding alignment and aggregation.

Findings

CMMI achieves accurate matrix recovery with minimal overlaps.

The method provides theoretical error bounds and normal approximations.

Simulation and real data show CMMI is efficient and effective.

Abstract

Motivated by the increasing demand for multi-source data integration in various scientific fields, in this paper we study matrix completion in scenarios where the data exhibits certain block-wise missing structures -- specifically, where only a few noisy submatrices representing (overlapping) parts of the full matrix are available. We propose the Chain-linked Multiple Matrix Integration (CMMI) procedure to efficiently combine the information that can be extracted from these individual noisy submatrices. CMMI begins by deriving entity embeddings for each observed submatrix, then aligns these embeddings using overlapping entities between pairs of submatrices, and finally aggregates them to reconstruct the entire matrix of interest. We establish, under mild regularity conditions, entrywise error bounds and normal approximations for the CMMI estimates. Simulation studies and real data applications show that CMMI is computationally efficient and effective in recovering the full matrix, even when overlaps between the observed submatrices are minimal.