Optimization and Systems Theory Seminar
Friday, February 6, 1998, 11.00-12.00, Room 3721, Lindstedtsvägen 25

Håkan Ekblom
Department of Mathematics
Luleå University of Technology

Algorithms for M-estimates

When fitting a model function to data, the least squares criterion is traditionally used. Unfortunately a least squares estimate is very sensitive= to errors in data. During the last three decades there has been a lot of intere= st in finding good alternatives to least squares. M-estimates, to where e.g. the Huber estimate belong, is an important class of estimates within so called robust regression.

I will give some statistical background to the computational problems considered. The main features of some algorithms designed to compute M-estimates will be presented. Also non-linear models will be treated.

Calendar of seminars
Last update: January 26, 1998 by Anders Forsgren, andersf@math.kth.se.