Show simple item record

Scaling Limit: Exact and Tractable Analysis of Online Learning Algorithms with Applications to Regularized Regression and PCA

dc.contributor.author Mattingly, Jonathan
dc.contributor.author Wang, Chuang
dc.contributor.author Lu, Yue M
dc.date.accessioned 2019-03-02T16:54:50Z
dc.date.available 2019-03-02T16:54:50Z
dc.identifier.uri https://hdl.handle.net/10161/18131
dc.description.abstract We present a framework for analyzing the exact dynamics of a class of online learning algorithms in the high-dimensional scaling limit. Our results are applied to two concrete examples: online regularized linear regression and principal component analysis. As the ambient dimension tends to infinity, and with proper time scaling, we show that the time-varying joint empirical measures of the target feature vector and its estimates provided by the algorithms will converge weakly to a deterministic measured-valued process that can be characterized as the unique solution of a nonlinear PDE. Numerical solutions of this PDE can be efficiently obtained. These solutions lead to precise predictions of the performance of the algorithms, as many practical performance metrics are linear functionals of the joint empirical measures. In addition to characterizing the dynamic performance of online learning algorithms, our asymptotic analysis also provides useful insights. In particular, in the high-dimensional limit, and due to exchangeability, the original coupled dynamics associated with the algorithms will be asymptotically "decoupled", with each coordinate independently solving a 1-D effective minimization problem via stochastic gradient descent. Exploiting this insight for nonconvex optimization problems may prove an interesting line of future research.
dc.subject cs.LG
dc.subject cs.LG
dc.subject cs.IT
dc.subject math.IT
dc.subject math.PR
dc.subject stat.ML
dc.title Scaling Limit: Exact and Tractable Analysis of Online Learning Algorithms with Applications to Regularized Regression and PCA
dc.type Journal article
dc.date.updated 2019-03-02T16:54:49Z
pubs.organisational-group Trinity College of Arts & Sciences
pubs.organisational-group Duke
pubs.organisational-group Mathematics
pubs.organisational-group Statistical Science
duke.contributor.orcid Mattingly, Jonathan|0000-0002-1819-729X


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record