pyrcc

A python implementation of Robust Continuous Clustering.

Sklearn style demonstration:

pyrcc

RCC is a clustering method introduced here: http://www.pnas.org/content/early/2017/08/28/1700770114

This is a port of the matlab implementation provided by the authors.

The code is self-contained in rcc.py

The following parameters are used in RCC:

  • k: (int)(deafult 10) number of neighbors used in the mutual KNN graph
  • verbose: (bool)(default True) verbosity
  • preprocessing: (string)(default "none") one of 'scale', 'minmax', 'normalization', 'none'. How to preprocess the features
  • measure: (string)(default "euclidean") one of 'cosine' or 'euclidean'. Paper used 'cosine'. Metric to use in constructing the mutual KNN graph
  • clustering_threshold: (float)(default 1.0) controls how agressively to assign points to clusters.

A demonstration of how to use this is shown in demo.py, measuring the AMI (adjusted mutual information) using the pendigits dataset.

GitHub