Principial Component Analysis (PCA) can be used to reduce the dimensionality of the high-dimensional data to simplify further analysis. For the
thesis the PCA was first used when analyzing all of the sequences and then to reduce the dimensionality of the median data prior to training the SOM with it.
The following Figures are based on the eigenvectors obtained from all the sequences.
 |
|
Figure 1: Comparision between the 1200-dimensional data space and the 80-dimensional eigenspace. On the left side are the original images, which are first projected into the 80-dimensional eigenspace and then reconstructed (decompressed) using the eigenvectors in the 1200-dimensional data space.
|
|
|
Figure 2: The first 20 eigenvectors with the highest eigenvalues.
|
