'Kernel' principal component analysis (PCA) is an elegant nonlinear generalisation of the popular linear data analysis method, where a kernel function implicitly defines a nonlinear transformation into a feature space wherein standard PCA is performed. Unfortunately, the technique is not 'sparse', since the components thus obtained are expressed in terms of kernels associated with every training vector. This paper shows that by approximating the covariance matrix in feature space by a reduced number of example vectors, using a maximum-likelihood approach, we may obtain a highly sparse form of kernel PCA without loss of effectiveness.
[1]
Yoshua Bengio,et al.
Pattern Recognition and Neural Networks
,
1995
.
[2]
Michael E. Tipping.
The Relevance Vector Machine
,
1999,
NIPS.
[3]
Michael E. Tipping,et al.
Probabilistic Principal Component Analysis
,
1999
.
[4]
Bernhard Schölkopf,et al.
Nonlinear Component Analysis as a Kernel Eigenvalue Problem
,
1998,
Neural Computation.
[5]
Dorothy T. Thayer,et al.
EM algorithms for ML factor analysis
,
1982
.
[6]
Shaogang Gong,et al.
A Multi-View Nonlinear Active Shape Model Using Kernel PCA
,
1999,
BMVC.