Faster Eigenvector Computation via Shift-and-Invert Preconditioning

Dan Garber, Elad Hazan, Chi Jin,  Sham, Cameron Musco, Praneeth Netrapalli, Aaron Sidford
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:2626-2634, 2016.

Abstract

We give faster algorithms and improved sample complexities for the fundamental problem of estimating the top eigenvector. Given an explicit matrix A∈Rn×d, we show how to compute an ϵ-approximate top eigenvector of ATA in time ˜O([nnz(A)+dsr(A)gap2]⋅log1/ϵ). Here nnz(A) is the number of nonzeros in A, sr(A) is the stable rank, and gap is the relative eigengap. We also consider an online setting in which, given a stream of i.i.d. samples from a distribution D with covariance matrix Σ and a vector x0 which is an O(gap) approximate top eigenvector for Σ, we show how to refine x0 to an ϵ approximation using O(var(D)gap−ϵ) samples from D. Here var(D) is a natural notion of variance. Combining our algorithm with previous work to initialize x0, we obtain improved sample complexities and runtimes under a variety of assumptions on D. We achieve our results via a robust analysis of the classic shift-and-invert preconditioning method. This technique lets us reduce eigenvector computation to approximately solving a series of linear systems with fast stochastic gradient methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v48-garber16, title = {Faster Eigenvector Computation via Shift-and-Invert Preconditioning}, author = {Garber, Dan and Hazan, Elad and Jin, Chi and Sham, and Musco, Cameron and Netrapalli, Praneeth and Sidford, Aaron}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {2626--2634}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/garber16.pdf}, url = {https://proceedings.mlr.press/v48/garber16.html}, abstract = {We give faster algorithms and improved sample complexities for the fundamental problem of estimating the top eigenvector. Given an explicit matrix $A \in \mathbb{R}^{n \times d}$, we show how to compute an $\epsilon$-approximate top eigenvector of $A^TA$ in time $\tilde O\left( \left[\text{nnz}(A) + \frac{d \text{sr}(A)}{\text{gap}^2} \right] \cdot \log 1/\epsilon\right)$. Here $\text{nnz}(A)$ is the number of nonzeros in $A$, $\text{sr}(A)$ is the stable rank, and gap is the relative eigengap. We also consider an online setting in which, given a stream of i.i.d. samples from a distribution D with covariance matrix $\Sigma$ and a vector $x_0$ which is an $O(\text{gap})$ approximate top eigenvector for $\Sigma$, we show how to refine $x_0$ to an $\epsilon$ approximation using $O \left( \frac{\text{var}(\mathcal{D})}{\text{gap}-\epsilon}\right)$ samples from $\mathcal{D}$. Here $\text{var}(\mathcal{D})$ is a natural notion of variance. Combining our algorithm with previous work to initialize $x_0$, we obtain improved sample complexities and runtimes under a variety of assumptions on D. We achieve our results via a robust analysis of the classic shift-and-invert preconditioning method. This technique lets us reduce eigenvector computation to approximately solving a series of linear systems with fast stochastic gradient methods.} }
Endnote
%0 Conference Paper %T Faster Eigenvector Computation via Shift-and-Invert Preconditioning %A Dan Garber %A Elad Hazan %A Chi Jin %A Sham %A Cameron Musco %A Praneeth Netrapalli %A Aaron Sidford %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-garber16 %I PMLR %P 2626--2634 %U https://proceedings.mlr.press/v48/garber16.html %V 48 %X We give faster algorithms and improved sample complexities for the fundamental problem of estimating the top eigenvector. Given an explicit matrix $A \in \mathbb{R}^{n \times d}$, we show how to compute an $\epsilon$-approximate top eigenvector of $A^TA$ in time $\tilde O\left( \left[\text{nnz}(A) + \frac{d \text{sr}(A)}{\text{gap}^2} \right] \cdot \log 1/\epsilon\right)$. Here $\text{nnz}(A)$ is the number of nonzeros in $A$, $\text{sr}(A)$ is the stable rank, and gap is the relative eigengap. We also consider an online setting in which, given a stream of i.i.d. samples from a distribution D with covariance matrix $\Sigma$ and a vector $x_0$ which is an $O(\text{gap})$ approximate top eigenvector for $\Sigma$, we show how to refine $x_0$ to an $\epsilon$ approximation using $O \left( \frac{\text{var}(\mathcal{D})}{\text{gap}-\epsilon}\right)$ samples from $\mathcal{D}$. Here $\text{var}(\mathcal{D})$ is a natural notion of variance. Combining our algorithm with previous work to initialize $x_0$, we obtain improved sample complexities and runtimes under a variety of assumptions on D. We achieve our results via a robust analysis of the classic shift-and-invert preconditioning method. This technique lets us reduce eigenvector computation to approximately solving a series of linear systems with fast stochastic gradient methods.
RIS
TY - CPAPER TI - Faster Eigenvector Computation via Shift-and-Invert Preconditioning AU - Dan Garber AU - Elad Hazan AU - Chi Jin AU - Sham AU - Cameron Musco AU - Praneeth Netrapalli AU - Aaron Sidford BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-garber16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 2626 EP - 2634 L1 - http://proceedings.mlr.press/v48/garber16.pdf UR - https://proceedings.mlr.press/v48/garber16.html AB - We give faster algorithms and improved sample complexities for the fundamental problem of estimating the top eigenvector. Given an explicit matrix $A \in \mathbb{R}^{n \times d}$, we show how to compute an $\epsilon$-approximate top eigenvector of $A^TA$ in time $\tilde O\left( \left[\text{nnz}(A) + \frac{d \text{sr}(A)}{\text{gap}^2} \right] \cdot \log 1/\epsilon\right)$. Here $\text{nnz}(A)$ is the number of nonzeros in $A$, $\text{sr}(A)$ is the stable rank, and gap is the relative eigengap. We also consider an online setting in which, given a stream of i.i.d. samples from a distribution D with covariance matrix $\Sigma$ and a vector $x_0$ which is an $O(\text{gap})$ approximate top eigenvector for $\Sigma$, we show how to refine $x_0$ to an $\epsilon$ approximation using $O \left( \frac{\text{var}(\mathcal{D})}{\text{gap}-\epsilon}\right)$ samples from $\mathcal{D}$. Here $\text{var}(\mathcal{D})$ is a natural notion of variance. Combining our algorithm with previous work to initialize $x_0$, we obtain improved sample complexities and runtimes under a variety of assumptions on D. We achieve our results via a robust analysis of the classic shift-and-invert preconditioning method. This technique lets us reduce eigenvector computation to approximately solving a series of linear systems with fast stochastic gradient methods. ER -
APA
Garber, D., Hazan, E., Jin, C., Sham, , Musco, C., Netrapalli, P. & Sidford, A.. (2016). Faster Eigenvector Computation via Shift-and-Invert Preconditioning. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:2626-2634 Available from https://proceedings.mlr.press/v48/garber16.html.

Related Material