Multilinear Mixing Model

Multilinear mixing models analyze data with multiple interacting factors, aiming to decompose complex datasets into simpler, interpretable components. Current research focuses on improving identifiability of these components, particularly in unpaired multimodal data, and developing efficient algorithms like those based on variable projection and tensor factorization for optimal representation and scalability. These models find applications in diverse fields, including hyperspectral imaging, multitask learning, and forecasting of multidimensional time series, offering improved accuracy and interpretability compared to traditional methods.