Optimal transform formed by a combination of nonlinear operators : the case of data dimensionality reduction
Date
2006
Authors
Torokhti, A.
Howlett, P.G.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Journal article
Citation
IEEE Transactions on Signal Processing, 2006; 54(4):1431-1444
Statement of Responsibility
Conference Name
Abstract
In this paper, a new approach to constructing optimal nonlinear transforms of random vectors is proposed and justified. The proposed transform T<inf>p</inf> is presented in the form of a sum with p terms where each term is interpreted as a particular rank-reduced transform. Moreover, terms in T<inf>p</inf> are represented as a combination of three operations F<inf>k</inf>, Q<inf>k</inf>, and φ<inf>k</inf> with k = 1,...,p. The prime idea is to determine F<inf>k</inf> separately, for each k = 1,...,p, from an associated rank-constrained minimization problem similar to that used in the Karhunen-Loève transform (KLT). The operations Q<inf>k</inf> and φ<inf>k</inf> are auxiliary for finding F<inf>k</inf>. The contribution of each term in T<inf>p</inf> improves the entire transform performance. A rigorous analysis of errors associated with the proposed transforms is given. It is shown that the proposed transform improves such characteristics of rank-reduced transforms as compression ratio and the accuracy of decompression and reduces required computational work. The Fourier series in Hilbert space, the Wiener filter, the KLT and the transforms presented in earlier author papers in the field are particular cases of the proposed method. Theoretical results are illustrated with numerical simulations. © 2006 IEEE.
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
Copyright status unknown