Home / Events / First-order methods for large scale optimisation problems – Modern first-order methods for structured convex problems
CCIMI Seminars

First-order methods for large scale optimisation problems – Modern first-order methods for structured convex problems

With Stephen Becker, University of Colorado

First-order methods for large scale optimisation problems – Modern first-order methods for structured convex problems

Part of the CCIMI short course ‘First-order methods for large scale optimisation problems’

Instructor: Stephen Becker, University of Colorado

Lecture 3: Modern first-order methods for structured convex problems
Starting with gradient descent, the proximal point method, and subgradient descent, we then explore accelerated and proximal variants (such as FISTA ), and finally discuss stochastic gradient descent (SGD) and variance-reduced versions such as SVRG and SAGA . There will be an interlude with convergence analysis for some of these methods and discussion of step-sizes.

Add to your calendar or Include in your list