Linear Discriminant Analysis Filter
Reduces the dimensionality of data via the Linear Discriminant Analysis.
Category |
|
Node |
|
Parameters |
Dimensionality: the number of dimensions to reduce the data to. This must be at most the number of applicable groups with training data minus one. Applicable (group-specific): whether a given group is applicable for this filter |
Inputs |
Input: the high-dimensional input data |
Outputs |
Output: the data projected onto the subspace |
Effect of the Filter
Similar to how the Principal Component Analysis works this filter reduces the dimensionality of a given set of input data. It attempts to find a subspace projection such that all groups that are applicable to this filter are separated as best as possible. For the determination of the subspace it will only consider the input data points (pixels) that were assigned to the applicable groups of the filter. (Non-assigned pixels or pixels that were assigned to groups that are not applicable to this filter will be ignored for the determination of the projection.)
After the projection has been found it will be applied to all input data of this filter.