-

 

 

 




Optimization Online





 

Fast and Faster Convergence of SGD for Over-Parameterized Models and an Accelerated Perceptron

Sharan Vaswani(sharanv***at***cs.ubc.ca)
Francis Bach(francis.bach***at***inria.fr)
Mark Schmidt(schmidtm***at***cs.ubc.ca)

Abstract: Modern machine learning focuses on highly expressive models that are able to fit or interpolate the data completely, resulting in zero training loss. For such models, we show that the stochastic gradients of common loss functions satisfy a strong growth condition. Under this condition, we prove that constant step-size stochastic gradient descent (SGD) with Nesterov acceleration matches the convergence rate of the deterministic accelerated method for both convex and strongly-convex functions. We also show that this condition implies that SGD can find a first-order stationary point as efficiently as full gradient descent in non-convex settings. Under interpolation, we further show that all smooth loss functions with a finite-sum structure satisfy a weaker growth condition. Given this weaker condition, we prove that SGD with a constant step-size attains the deterministic convergence rate in both the strongly-convex and convex settings. Under additional assumptions, the above results enable us to prove an O(1/k^2) mistake bound for k iterations of a stochastic perceptron algorithm using the squared-hinge loss. Finally, we validate our theoretical findings with experiments on synthetic and real datasets.

Keywords: Stochastic gradient descent, Nesterov acceleration, Over-parametrization, Interpolation

Category 1: Convex and Nonsmooth Optimization (Convex Optimization )

Category 2: Convex and Nonsmooth Optimization (Generalized Convexity/Monoticity )

Citation:

Download: [PDF]

Entry Submitted: 02/25/2019
Entry Accepted: 02/25/2019
Entry Last Modified: 02/25/2019

Modify/Update this entry


  Visitors Authors More about us Links
  Subscribe, Unsubscribe
Digest Archive
Search, Browse the Repository

 

Submit
Update
Policies
Coordinator's Board
Classification Scheme
Credits
Give us feedback
Optimization Journals, Sites, Societies
Mathematical Optimization Society