Convergence of Stochastic Gradient Descent for PCA
This addresses a theoretical challenge in non-convex optimization for machine learning, offering incremental improvements in convergence analysis for PCA.
The paper tackles the problem of analyzing stochastic gradient descent (SGD) for principal component analysis (PCA) in a streaming stochastic setting, providing the first eigengap-free convergence guarantees and, under an eigengap assumption, new guarantees with improved dependence on the eigengap.
We consider the problem of principal component analysis (PCA) in a streaming stochastic setting, where our goal is to find a direction of approximate maximal variance, based on a stream of i.i.d. data points in $\reals^d$. A simple and computationally cheap algorithm for this is stochastic gradient descent (SGD), which incrementally updates its estimate based on each new data point. However, due to the non-convex nature of the problem, analyzing its performance has been a challenge. In particular, existing guarantees rely on a non-trivial eigengap assumption on the covariance matrix, which is intuitively unnecessary. In this paper, we provide (to the best of our knowledge) the first eigengap-free convergence guarantees for SGD in the context of PCA. This also partially resolves an open problem posed in \cite{hardt2014noisy}. Moreover, under an eigengap assumption, we show that the same techniques lead to new SGD convergence guarantees with better dependence on the eigengap.