A consistent and flexible framework for deep matrix factorizations
This work addresses a methodological inconsistency in unsupervised data mining for researchers in machine learning and data analysis, though it appears incremental as it builds on existing deep MF techniques.
The paper tackled the inconsistency in loss functions across layers in deep matrix factorizations by introducing two new loss functions and a generic optimization framework, demonstrating effectiveness with constraints like sparsity and nonnegativity on synthetic and real data for tasks such as hyperspectral unmixing and facial feature extraction.
Deep matrix factorizations (deep MFs) are recent unsupervised data mining techniques inspired by constrained low-rank approximations. They aim to extract complex hierarchies of features within high-dimensional datasets. Most of the loss functions proposed in the literature to evaluate the quality of deep MF models and the underlying optimization frameworks are not consistent because different losses are used at different layers. In this paper, we introduce two meaningful loss functions for deep MF and present a generic framework to solve the corresponding optimization problems. We illustrate the effectiveness of this approach through the integration of various constraints and regularizations, such as sparsity, nonnegativity and minimum-volume. The models are successfully applied on both synthetic and real data, namely for hyperspectral unmixing and extraction of facial features.