A1799
Title: Representation learning with blockwise missingness and signal heterogeneity
Authors: Weijing Tang - Carnegie Mellon University (United States) [presenting]
Abstract: Unified representation learning for multi-source data integration faces two important challenges: blockwise missingness and blockwise signal heterogeneity. The former arises from sources observing different, yet potentially overlapping, feature sets, while the latter involves varying signal strengths across subject groups and feature sets. While existing methods perform well with fully observed data or uniform signal strength, their performance degenerates when these two challenges coincide, which is common in practice. To address this, Anchor Projected Principal Component Analysis (APPCA) is proposed as a general framework for representation learning with structured blockwise missingness that is robust to signal heterogeneity. APPCA first recovers robust group-specific column spaces using all observed feature sets, and then aligns them by projecting shared Anchor features onto these subspaces before performing PCA. Estimation error bounds for embedding reconstruction are established through a fine-grained perturbation Analysis. In particular, using a novel spectral slicing technique, the bound eliminates the standard dependency on the signal strength of subject embeddings, relying instead solely on the signal strength of integrated feature sets. The proposed method is validated through extensive simulation studies and an application to multimodal single-cell sequencing data.
