fastcluster

To help you get started, we've collected the most common ways that fastcluster is being used within popular public projects.

mwaskom / seaborn / seaborn / matrix.py View on Github

def _calculate_linkage_fastcluster(self):
        import fastcluster
        # Fastcluster has a memory-saving vectorized version, but only
        # with certain linkage methods, and mostly with euclidean metric
        vector_methods = ('single', 'centroid', 'median', 'ward')
        euclidean_methods = ('centroid', 'median', 'ward')
        euclidean = self.metric == 'euclidean' and self.method in \
            euclidean_methods
        if euclidean or self.method == 'single':
            return fastcluster.linkage_vector(self.array,
                                              method=self.method,
                                              metric=self.metric)
        else:
            pairwise_dists = distance.pdist(self.array, metric=self.metric)
            linkage = fastcluster.linkage(pairwise_dists, method=self.method)
            del pairwise_dists
            return linkage

View more ways to use fastcluster

This library provides Python functions for hierarchical clustering. It generates hierarchical clusters from distance matrices or from vector data.

This module is intended to replace the functions

    linkage, single, complete, average, weighted, centroid, median, ward

in the module scipy.cluster.hierarchy with the same functionality but much faster algorithms. Moreover, the function linkage_vector provides memory-efficient clustering for vector data.

The interface is very similar to MATLAB's Statistics Toolbox API to make code easier to port from MATLAB to Python/NumPy. The core implementation of this library is in C++ for efficiency.

User manual: fastcluster.pdf.

The “Yule” distance function changed in fastcluster version 1.2.0. This is following a change in SciPy 1.6.3. It is recommended to use fastcluster version 1.1.x together with SciPy versions before 1.6.3 and fastcluster 1.2.x with SciPy ≥1.6.3.

The fastcluster package is considered stable and will undergo few changes from now on. If some years from now there have not been any updates, this does not necessarily mean that the package is unmaintained but maybe it just was not necessary to correct anything. Of course, please still report potential bugs and incompatibilities to daniel@danifold.net. You may also use my GitHub repository for bug reports, pull requests etc.

Note that PyPI and my GitHub repository host the source code for the Python interface only. The archive with both the R and the Python interface is available on CRAN and the GitHub repository “cran/fastcluster”. Even though I appear as the author also of this second GitHub repository, this is just an automatic, read-only mirror of the CRAN archive, so please do not attempt to report bugs or contact me via this repository.

Installation files for Windows are provided on PyPI and on Christoph Gohlke's web page.

Christoph Dalitz wrote a pure C++ interface to fastcluster.

Reference: Daniel Müllner, fastcluster: Fast Hierarchical, Agglomerative Clustering Routines for R and Python, Journal of Statistical Software, 53 (2013), no. 9, 1–18, https://doi.org/10.18637/jss.v053.i09.