Principal Component Analysis With Complex Kernel: The Widely Linear Model
Authors:
Open Access
- 28 October 2013
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 25 (9) , 1719-1726
Abstract
Nonlinear complex representations, via the use of complex kernels, can be applied to model and capture the nonlinearities of complex data. Even though the theoretical tools of complex reproducing kernel Hilbert spaces (CRKHS) have been recently successfully applied to the design of digital filters and regression and classification frameworks, there is a limited research on component analysis and dimensionality reduction in CRKHS. The aim of this brief is to properly formulate the most popular component analysis methodology, i.e., Principal Component Analysis (PCA), in CRKHS. In particular, we define a general widely linear complex kernel PCA framework. Furthermore, we show how to efficiently perform widely linear PCA in small sample sized problems. Finally, we show the usefulness of the proposed framework in robust reconstruction using Euler data representation.Keywords
Funding Information
- Act ¿Scholarships programme SSF (State Scholarships Foundation/IKY) with an individualized assessment process of the Academic Year 2011-2012¿ from resources of the Operational Programme ¿Education and Lifelong Learning¿, of the European Social Fund (ESF),
- Engineering and Physical Sciences Research Council (EPSRC) project EP/J017787/1 Analysis of Facial Behaviour for Security in 4D (4D-FAB)
This publication has 35 references indexed in Scilit:
- Projection-Based Fast Learning Fully Complex-Valued Relaxation Neural NetworkIEEE Transactions on Neural Networks and Learning Systems, 2013
- Generalization Characteristics of Complex-Valued Feedforward Neural Networks in Relation to Signal CoherenceIEEE Transactions on Neural Networks and Learning Systems, 2012
- Subspace analysis of arbitrarily many linear filter responses with an application to face tracking2011
- Robust Principal Component Analysis Based on Maximum Correntropy CriterionIEEE Transactions on Image Processing, 2011
- Statistical Signal Processing of Complex-Valued Data2010
- Principal Component Analysis Based on L1-Norm MaximizationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2008
- An Explicit Description of the Reproducing Kernel Hilbert Spaces of Gaussian RBF KernelsIEEE Transactions on Information Theory, 2006
- R1-PCA2006
- The Pre-Image Problem in Kernel MethodsIEEE Transactions on Neural Networks, 2004
- Widely linear estimation with complex dataIEEE Transactions on Signal Processing, 1995