Principal components analysis plugin for the ManiVault visual analytics framework.
Settings UI (left) and several PCA components of MNIST (right: 1&2, 3&4, 5&6)
Clone the repo:
git clone https://github.com/ManiVaultStudio/PcaPlugin.git
Eigen is automatically downloaded during the cmake configuration.
- Preprocessing:
- By default, the plugin internally centers the data so that each dimension/channel has zero mean.
- Optional normalization steps before this centering: Mean normalization and Rescaling (min-max normalization).
- PCA computation algorithms (implemented with Eigen):
- Explicitly computing the eigenvectors of the covariance matrix
- Singular value decomposition
- Number of components:
- Defaults to two
You can perform unit tests. Set the cmake variable MV_PCA_UNIT_TESTS to build tests. To build the testing project, you'll need to install some further dependencies and create ground truth data; see test/README.md.