Title :
Topology-based genetic search for the Stahel-Donoho estimator
Author :
Muruzábal, Jorge
fDate :
Nov. 29 1995-Dec. 1 1995
Abstract :
Woodruff and Rocke (1993) have shown the usefulness of genetic searching for robust multivariate estimation via the minimum-volume ellipsoid estimator. In this paper, the topology-based scheme proposed by Davidor et al. (1993) is investigated in a similar vein when the goal is to approximate the Stahel-Donoho estimator. The main interest of this work relies on the fact that no strategy other than a random search is currently known for the computation of the latter estimator. Besides determining the best genetic strategy, it is also worth while trying to rate different prototypical patterns for outlying data in terms of their intrinsic difficulty
Keywords :
Covariance matrix; Degradation; Econometrics; Ellipsoids; Genetics; Maximum likelihood estimation; Prototypes; Robustness; Statistics; Veins;
Conference_Titel :
Evolutionary Computation, 1995., IEEE International Conference on
Conference_Location :
Perth, WA, Australia
Print_ISBN :
0-7803-2759-4
DOI :
10.1109/ICEC.1995.489131