• Quick start
• What's included
• Getting Started
• Creators
• Visualizations

Get started by downloading the latest release:

• Download the latest release
• Clone the repo: git clone https://github.com/NicoMigenda/NGPCA-Clustering.git

Within the download you'll find the following directories and files:

|-- Example_stationary.m
|-- Example_stationary.mlx
|-- README.md
|-- Results
|   `-- gif
|       |-- s1_G_AR_S_V.gif
|       |-- s2_G_AR_S_V.gif
|       |-- s3_G_AR_S_V.gif
|       `-- s4_G_AR_S_V.gif
|-- data
|   `-- s1.mat
`-- ngpca
  |-- NGPCA.m
  |-- drawunits.m
  |-- eforrlsa.m
  |-- init.m
  |-- normalizedmi.m
  |-- plot_ellipse.m
  |-- potentialFunction.m
  |-- update.m
  |-- validate_CI.m
  `-- validate_NMI_DU.m

The latest release contains all files needed to directly run the algorithm:

1. Open Example.m or Example.mlx in Matlab
2. Compiling the script will automatically perform NGPCA Clustering on the S1 data set with standard settings

Optional:

3. Change default settings or add optional parameters to the ngpca object creation or for the training process
4. Train the model directly on a full data set using the fit_multiple() function or build your own training loops with fit_single()
5. Visualize the clustering results with the draw() function
6. Calculate validation metrics (CI, NMI, DU) by provding ground thruth and cluster shape information

The following visualizations represent the learning process on selected data sets of the standard clustering benchmark database. For all data sets the default settings are used.

Nico Migenda

Ralf Möller

Wolfram Schenck