Accelerated Factored Gradient Descent for Low-Rank Matrix Factorization

Dongruo Zhou, Yuan Cao, Quanquan Gu
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:4430-4440, 2020.

Abstract

We study the low-rank matrix estimation problem, where the objective function $\mathcal{L}(\Mb)$ is defined over the space of positive semidefinite matrices with rank less than or equal to $r$. A fast approach to solve this problem is matrix factorization, which reparameterizes $\mathbf{M}$ as the product of two smaller matrix such that $\mathbf{M} =\mathbf{U}\mathbf{U}^\top$ and then performs gradient descent on $\mathbf{U}$ directly, a.k.a., factored gradient descent. Since the resulting problem is nonconvex, whether Nesterov’s acceleration scheme can be adapted to it remains a long-standing question. In this paper, we answer this question affirmatively by proposing a novel and practical accelerated factored gradient descent method motivated by Nesterov’s accelerated gradient descent. The proposed method enjoys better iteration complexity and computational complexity than the state-of-the-art algorithms in a wide regime. The key idea of our algorithm is to restrict all its iterates onto a special convex set, which enables the acceleration. Experimental results demonstrate the faster convergence of our algorithm and corroborate our theory.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-zhou20b, title = {Accelerated Factored Gradient Descent for Low-Rank Matrix Factorization}, author = {Zhou, Dongruo and Cao, Yuan and Gu, Quanquan}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {4430--4440}, year = {2020}, editor = {Silvia Chiappa and Roberto Calandra}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/zhou20b/zhou20b.pdf}, url = { http://proceedings.mlr.press/v108/zhou20b.html }, abstract = {We study the low-rank matrix estimation problem, where the objective function $\mathcal{L}(\Mb)$ is defined over the space of positive semidefinite matrices with rank less than or equal to $r$. A fast approach to solve this problem is matrix factorization, which reparameterizes $\mathbf{M}$ as the product of two smaller matrix such that $\mathbf{M} =\mathbf{U}\mathbf{U}^\top$ and then performs gradient descent on $\mathbf{U}$ directly, a.k.a., factored gradient descent. Since the resulting problem is nonconvex, whether Nesterov’s acceleration scheme can be adapted to it remains a long-standing question. In this paper, we answer this question affirmatively by proposing a novel and practical accelerated factored gradient descent method motivated by Nesterov’s accelerated gradient descent. The proposed method enjoys better iteration complexity and computational complexity than the state-of-the-art algorithms in a wide regime. The key idea of our algorithm is to restrict all its iterates onto a special convex set, which enables the acceleration. Experimental results demonstrate the faster convergence of our algorithm and corroborate our theory. } }
Endnote
%0 Conference Paper %T Accelerated Factored Gradient Descent for Low-Rank Matrix Factorization %A Dongruo Zhou %A Yuan Cao %A Quanquan Gu %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-zhou20b %I PMLR %P 4430--4440 %U http://proceedings.mlr.press/v108/zhou20b.html %V 108 %X We study the low-rank matrix estimation problem, where the objective function $\mathcal{L}(\Mb)$ is defined over the space of positive semidefinite matrices with rank less than or equal to $r$. A fast approach to solve this problem is matrix factorization, which reparameterizes $\mathbf{M}$ as the product of two smaller matrix such that $\mathbf{M} =\mathbf{U}\mathbf{U}^\top$ and then performs gradient descent on $\mathbf{U}$ directly, a.k.a., factored gradient descent. Since the resulting problem is nonconvex, whether Nesterov’s acceleration scheme can be adapted to it remains a long-standing question. In this paper, we answer this question affirmatively by proposing a novel and practical accelerated factored gradient descent method motivated by Nesterov’s accelerated gradient descent. The proposed method enjoys better iteration complexity and computational complexity than the state-of-the-art algorithms in a wide regime. The key idea of our algorithm is to restrict all its iterates onto a special convex set, which enables the acceleration. Experimental results demonstrate the faster convergence of our algorithm and corroborate our theory.
APA
Zhou, D., Cao, Y. & Gu, Q.. (2020). Accelerated Factored Gradient Descent for Low-Rank Matrix Factorization. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:4430-4440 Available from http://proceedings.mlr.press/v108/zhou20b.html .

Related Material