We introduce a greedy algorithm that works from coarse to fine by iteratively applying localized principal component analysis (PCA). The decision where and when to split or add new components is based on two antagonistic criteria. Firstly, the well known quadratic reconstruction error and secondly a measure for the homogeneity of the distribution. For the latter criterion, which we call “generation error”, we compared two different possible methods to assess if the data samples are distributed homogeneously. The proposed algorithm does not involve a costly multi-objective optimization to find a partition of the inputs. Further, the final number of local PCA units, as well as their individual dimensionality need not to be predefined. We demonstrate that the method can flexibly react to different intrinsic dimensionalities of the data.
[1]
Nathan Intrator,et al.
Learning as Extraction of Low-Dimensional Representations
,
1997
.
[2]
Teuvo Kohonen,et al.
The self-organizing map
,
1990
.
[3]
S T Roweis,et al.
Nonlinear dimensionality reduction by locally linear embedding.
,
2000,
Science.
[4]
H. Ritter,et al.
Local PCA learning with resolution-dependent mixtures of Gaussians
,
1999
.
[5]
Heiko Hoffmann,et al.
An extension of neural gas to local PCA
,
2004,
Neurocomputing.
[6]
Nanda Kambhatla,et al.
Dimension Reduction by Local Principal Component Analysis
,
1997,
Neural Computation.
[7]
Matthew Brand,et al.
Charting a Manifold
,
2002,
NIPS.
[8]
D. Cox.
Karl Pearson and the Chi-Squared Test
,
2002
.