Dropout as a Low-Rank Regularizer for Matrix Factorization

Jacopo Cavazza, Pietro Morerio, Benjamin Haeffele, Connor Lane, Vittorio Murino, Rene Vidal
Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, PMLR 84:435-444, 2018.

Abstract

Regularization for matrix factorization (MF) and approximation problems has been carried out in many different ways. Due to its popularity in deep learning, dropout has been applied also for this class of problems. Despite its solid empirical performance, the theoretical properties of dropout as a regularizer remain quite elusive for this class of problems. In this paper, we present a theoretical analysis of dropout for MF, where Bernoulli random variables are used to drop columns of the factors. We demonstrate the equivalence between dropout and a fully deterministic model for MF in which the factors are regularized by the sum of the product of squared Euclidean norms of the columns. Additionally, we inspect the case of a variable sized factorization and we prove that dropout achieves the global minimum of a convex approximation problem with (squared) nuclear norm regularization. As a result, we conclude that dropout can be used as a low-rank regularizer with data dependent singular-value thresholding.

Cite this Paper


BibTeX
@InProceedings{pmlr-v84-cavazza18a, title = {Dropout as a Low-Rank Regularizer for Matrix Factorization}, author = {Cavazza, Jacopo and Morerio, Pietro and Haeffele, Benjamin and Lane, Connor and Murino, Vittorio and Vidal, Rene}, booktitle = {Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics}, pages = {435--444}, year = {2018}, editor = {Storkey, Amos and Perez-Cruz, Fernando}, volume = {84}, series = {Proceedings of Machine Learning Research}, month = {09--11 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v84/cavazza18a/cavazza18a.pdf}, url = {https://proceedings.mlr.press/v84/cavazza18a.html}, abstract = {Regularization for matrix factorization (MF) and approximation problems has been carried out in many different ways. Due to its popularity in deep learning, dropout has been applied also for this class of problems. Despite its solid empirical performance, the theoretical properties of dropout as a regularizer remain quite elusive for this class of problems. In this paper, we present a theoretical analysis of dropout for MF, where Bernoulli random variables are used to drop columns of the factors. We demonstrate the equivalence between dropout and a fully deterministic model for MF in which the factors are regularized by the sum of the product of squared Euclidean norms of the columns. Additionally, we inspect the case of a variable sized factorization and we prove that dropout achieves the global minimum of a convex approximation problem with (squared) nuclear norm regularization. As a result, we conclude that dropout can be used as a low-rank regularizer with data dependent singular-value thresholding.} }
Endnote
%0 Conference Paper %T Dropout as a Low-Rank Regularizer for Matrix Factorization %A Jacopo Cavazza %A Pietro Morerio %A Benjamin Haeffele %A Connor Lane %A Vittorio Murino %A Rene Vidal %B Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2018 %E Amos Storkey %E Fernando Perez-Cruz %F pmlr-v84-cavazza18a %I PMLR %P 435--444 %U https://proceedings.mlr.press/v84/cavazza18a.html %V 84 %X Regularization for matrix factorization (MF) and approximation problems has been carried out in many different ways. Due to its popularity in deep learning, dropout has been applied also for this class of problems. Despite its solid empirical performance, the theoretical properties of dropout as a regularizer remain quite elusive for this class of problems. In this paper, we present a theoretical analysis of dropout for MF, where Bernoulli random variables are used to drop columns of the factors. We demonstrate the equivalence between dropout and a fully deterministic model for MF in which the factors are regularized by the sum of the product of squared Euclidean norms of the columns. Additionally, we inspect the case of a variable sized factorization and we prove that dropout achieves the global minimum of a convex approximation problem with (squared) nuclear norm regularization. As a result, we conclude that dropout can be used as a low-rank regularizer with data dependent singular-value thresholding.
APA
Cavazza, J., Morerio, P., Haeffele, B., Lane, C., Murino, V. & Vidal, R.. (2018). Dropout as a Low-Rank Regularizer for Matrix Factorization. Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 84:435-444 Available from https://proceedings.mlr.press/v84/cavazza18a.html.

Related Material