Lieven Vandenberghe (UCLA)

Nov 13, 2017.

Title and Abstract

Primal-dual first-order methods for convex optimization

The talk will be on primal-dual first order methods derived from the Douglas-Rachford operator splitting algorithm. We will start with some applications to image deblurring problems that illustrate the versatility of the Douglas-Rachford method for primal-dual decomposition in large scale optimization. The second part of the talk will be concerned with the important primal-dual hybrid gradient (PDHG) method. This method is widely used in image processing, computer vision, and machine learning. We will show how the PDHG method can be derived from the Douglas-Rachford splitting method, and discuss some implications of this derivation for the convergence analysis and extensions of PDHG. (Joint work with Daniel O'Connor.)

Bio

Lieven Vandenberghe is Professor in the Electrical and Computer Engineering Department at the University of California, Los Angeles, with a joint appointment in the Department of Mathematics. He received a Ph.D. in Electrical Engineering from K.U. Leuven, Belgium, in 1992. He joined UCLA in 1997, following postdoctoral appointments at K.U. Leuven and Stanford University, and has held visiting professor positions at K.U. Leuven and the Technical University of Denmark. He is coauthor (with Stephen Boyd) of the book Convex Optimization (2004) and editor (with Henry Wolkowicz and Romesh Saigal) of the Handbook of Semidefinite Programming (2000)