RMITOpt Seminar – Andrew Eberhard (RMIT)

  Speaker:   Prof. Andrew Eberhard – RMIT University

Title:  Divide and Concur – An Overview and some examples

Date and Time:  Friday, April 5th, 3.00pm – 4.00pm

Location: Building 8 Level 9 Room 66 (AGR) RMIT City campus

Abstract:    Divide and concur is a play on words that deliberately echo computing science mantra of divide and conquer. In divide and conquer a problem is broken down into smaller and smaller subproblems until each smaller part is easily solved. In divide and concur the subproblems remain linked by a relax constraint that ultimately must be reconciled.

These ideas have found their way into feasibility problems and optimisation both continuous and discrete. Some view these are project-project feasibility problems, but recent work has suggested they are really an expression the older Gauss-Seidel approach.

In this talk we will discuss a modified version of a penalty-based Gauss-Seidel method as applied to stochastic optimisation that has stronger theoretical properties for certain sub-classes of Stochastic Integer Programs (SIP). In particular we will show that the modified algorithm always converges to a feasible point and this is verified by numerical experiments. This method shares structure and properties related the feasibility pump of integer programming and progressive Hedging heuristics used for stochastic integer programming.

Bio:    Andrew Eberhard did his PhD at Adelaide University under Prof. Charles Pearce and after graduating spend some time at UniSA (then SAIT) before moving to RMIT in Melbourne in the early 1990s. He has been an active member of the Australian mathematical and optimisation community for more than 20 years. He has served on the executives of ASOR, ANZIAM, as the deputy director of AMSI and the board of AMSI. Currently he is the co-chair of the AustMS special interest group Mathematics of Computation and Optimisation (MoCaO). His interests span numerous areas including nonsmooth and variational analysis, optimisation algorithms (both continuous and discrete), systems and control theory, operations research and other more theoretical aspect of optimisation theory.

 

Skip to toolbar