Robust independent component analysis algorithms for projection pursuit
Ruck Thawonmas, Jianting Cao · 2003
This paper presents the derivation of batch-mode neural algorithms which seek to robustly identify interesting projections of high dimensional data. The new index for projection pursuit, is a measure of the difference between a platykurtic density and a leptokurtic density. The robustness experiment is conducted to verify the validity of the proposed index when the algorithms are applied to artificial data and commonly used benchmark "crab" data.