Implementation of the Hungarian (Munkres) Algorithm using np. Usage: hungarian = Hungarian(cost_matrix) hungarian.calculate(

Basic implementation of Spherical K-Means (SKM), backed by a generalized K-Means implementation that uses cosine distances. Term rankings are

A top-weighted version of Jaccard, which takes into account rank positions. This is based on Fagin's Average Overlap Intersection Metric.

Use minimum number of lines to cover all zeros in the matrix. Algorithm based on: http://weber.ucsd.edu/~vcrawfor/hungar.pdf

Relevance function with inverse rank-weighted scores.

Validation measure that calculates the average pairwise dissimilarity between all unique pairs of term rankings in a given ranking set. The defa

Simple binary Jaccard-based ranking comparison, which does not take into account rank positions.

DCG-based rank relevance function, with inverse log-weighted scores.

Wrapper class for Mallet. Requires that a binary version of Mallet 2.0 is available locally.

Wrapper class backed by the Nimfa package NMF implementation.

A validator that evaluates topic (cluster) memberships based on an external set of ground truth classes. Note: we assume both the topic and class m

Calculates the agreement between pairs of ranking sets, using a specified measure of similarity between rankings.

Generic rank relevance function, with unit weights for all ranks.

A utility class for keeping track of experiment scores produced by multiple validation measures applied to different topic models.

Wrapper class backed by the scikit-learn package NMF implementation.

Validation measure, which compares the agreement between term rankings derived from the centroids produced using a 'ground truth' partition, with a