Subspace Projection Filter

Calculate the signal subspace for the given input data using the HySime algorithm.

Category

category_decomposition Decomposition

Node

subspaceidentification_node

Parameters

CutoffMethod: how to determine when the subspace dimension is sufficiently large:

  • Fixed: the subspace dimension is fixed

  • Automatic: automatically attempt to determine the subspace dimension

Dimensionality: how many dimensions the subspace should have (in the Fixed case)

Output Configuration: allows for enabling of the optional Basis output

Inputs

Input: the input data

Outputs

Output: the data projected onto the subspace

Basis (optional): training-only data that contains the basis vectors of the subspace

Effect of the Filter

This filter uses the HySime algorithm to determine the signal subspace of a given set of training data. It will record that subspace basis and project all input data during execution onto that subspace.

This may be used to perform a dimensionality reduction similar to what the Principal Component Analysis does.