Tuning-Free Online Robust Principal Component Analysis through Implicit Regularization
This work addresses the scalability issue for large datasets by eliminating the need for dataset-dependent parameter tuning in robust PCA, though it is incremental as it builds on existing OR-PCA methods.
The paper tackled the problem of dataset-sensitive tuning in Online Robust Principal Component Analysis (OR-PCA) by using implicit regularization from modified gradient descents to make it tuning-free, achieving comparable or better performance than tuned OR-PCA on simulated and real-world datasets.
The performance of the standard Online Robust Principal Component Analysis (OR-PCA) technique depends on the optimum tuning of the explicit regularizers and this tuning is dataset sensitive. We aim to remove the dependency on these tuning parameters by using implicit regularization. We propose to use the implicit regularization effect of various modified gradient descents to make OR-PCA tuning free. Our method incorporates three different versions of modified gradient descent that separately but naturally encourage sparsity and low-rank structures in the data. The proposed method performs comparable or better than the tuned OR-PCA for both simulated and real-world datasets. Tuning-free ORPCA makes it more scalable for large datasets since we do not require dataset-dependent parameter tuning.