Constrained Multimodal Data Mining Using Coupled Matrix and Tensor Factorizations

Explore advanced techniques for analyzing complex multimodal data sets through a conference talk on constrained multimodal data mining using coupled matrix and tensor factorizations. Delve into the challenges of extracting insights from heterogeneous data sources, such as dynamic metabolomics and static genetic information. Learn about the extension of tensor factorizations to joint analysis through coupled matrix and tensor factorizations (CMTF) and discover a flexible algorithmic framework based on Alternating Optimization (AO) and the Alternating Direction Method of Multipliers (ADMM). Examine the application of these methods to reveal underlying patterns, their evolution over time, and improved subject stratifications in complex systems like the human metabolome and brain. Gain insights into ongoing research from the TrACEr project and understand the potential of these techniques for advancing explainable AI in scientific domains.

Evrim Acar - Constrained Multimodal Data Mining using Coupled Matrix and Tensor Factorizations