Streaming PCA: Matching Matrix Bernstein and Near-Optimal Finite Sample Guarantees for Oja's Algorithm

Abstract

This work provides improved guarantees for streaming principle component analysis (PCA). Given $A_1,\dots ,A_n\in {\mathbb{R}}^{d\times d}$ sampled independently from distributions satisfying $\mathbb{E}[A_i]=\Sigma$ for $\Sigma ⪰0$, this work provides an $O(d)$-space linear-time single-pass streaming algorithm for estimating the top eigenvector of $\Sigma$. The algorithm nearly matches (and in certain cases improves upon) the accuracy obtained by the standard batch method that computes top eigenvector of the empirical covariance $\frac{1}{n}{\sum }_{i\in [n]}{A}_i$ as analyzed by the matrix Bernstein inequality. Moreover, to achieve constant accuracy, our algorithm improves upon the best previous known sample complexities of streaming algorithms by either a multiplicative factor of $O(d)$ or $1/\mathrm{gap}$ where $\mathrm{gap}$ is the relative distance between the top two eigenvalues of $\Sigma$. These results are achieved through a novel analysis of the classic Oja's algorithm, one of the oldest and most popular algorithms for streaming PCA. In particular, this work shows that simply picking a random initial point $w_0$ and applying the update rule $w_{i+1}=w_i+{\eta }_i{A}_i{w}_i$ suffices to accurately estimate the top eigenvector, with a suitable choice of ${\eta }_i$. We believe our result sheds light on how to efficiently perform streaming PCA both in theory and in practice and we hope that our analysis may serve as the basis for analyzing many variants and extensions of streaming PCA.

arXiv commentUpdated title

References in corpus (6)

• The Noisy Power Method: A Meta Algorithm with Applications
• Optimal approximate matrix product in terms of stable rank
• Fast and Simple PCA via Convex Optimization
• Robust Shift-and-Invert Preconditioning: Faster and More Sample Efficient Algorithms for Eigenvector Computation
• Frequent Directions : Simple and Deterministic Matrix Sketching
• Geometric Median in Nearly Linear Time

Cited by in corpus (8)

• Communication-efficient Algorithms for Distributed Stochastic Principal Component Analysis
• Fixed-Rank Approximation of a Positive-Semidefinite Matrix from Streaming Data
• Accelerated Stochastic Power Iteration
• History PCA: A New Algorithm for Streaming PCA
• Nearly Optimal Stochastic Approximation for Online Principal Subspace Estimation
• Smooth Sensitivity Based Approach for Differentially Private Principal Component Analysis
• On the Regret Minimization of Nonconvex Online Gradient Ascent for Online PCA
• A New Spectral Method for Latent Variable Models