Fast and Accurate Domain Adaptation for Irregular Tensor Decomposition

Junghun Kim, Ka Hyun Park, Jun-Gi Jang, U Kang

Abstract: Given an irregular tensor from a newly emerging domain, how can we quickly and accurately capture its patterns utilizing existing irregular tensors in multiple domains? The problem is of great importance for various tasks such as finding patterns of a new disease using pre-existing diseases data. This is challenging as new target tensors have limited information due to their recent emergence. Thus, carefully utilizing the existing source tensors for analyzing the target tensor is helpful. PARAFAC2 decomposition is a strong tool for finding the patterns of irregular tensors, and the patterns are used in many applications such as missing value prediction and anomaly detection. However, previous PARAFAC2-based works cannot adaptably handle newly emerging target tensors utilizing the source tensors.In this work, we propose Meta-P2, a fast and accurate domain adaptation method for irregular tensor decomposition. Meta-P2 generates a meta factor matrix from the multiple source domains, by domain adaptation and meta-update steps. Meta-P2 quickly and accurately finds the patterns of the new irregular tensor utilizing the meta factor matrix. Extensive experiments on real-world datasets show that Meta-P2 achieves the best performance in various downstream tasks including missing value prediction and anomaly detection tasks.