Regularized L21-Based Semi-NonNegative Matrix Factorization
This work addresses data compression challenges in machine learning for mixed-sign data, but it appears incremental as it builds on existing matrix factorization methods with specific regularization.
The paper tackles the problem of data compression for mixed-sign data requiring high-fidelity reconstruction, presenting Regularized L21 Semi-Non-Non-Negative Matrix Factorization (L21 SNF) as a robust, parts-based algorithm with a proof of convergence and demonstrated advantages in compressing overdetermined systems in machine learning.
We present a general-purpose data compression algorithm, Regularized L21 Semi-NonNegative Matrix Factorization (L21 SNF). L21 SNF provides robust, parts-based compression applicable to mixed-sign data for which high fidelity, individualdata point reconstruction is paramount. We derive a rigorous proof of convergenceof our algorithm. Through experiments, we show the use-case advantages presentedby L21 SNF, including application to the compression of highly overdeterminedsystems encountered broadly across many general machine learning processes.