sitator.landmark.cluster package¶

Submodules¶

Cluster landmark vectors using the custom online algorithm from the original paper.

sitator.landmark.cluster.dotprod.do_landmark_clustering(landmark_vectors, clustering_params, min_samples, verbose)¶

Cluster landmarks into sites using Markov Clustering and then assign each landmark vector.

Valid clustering params include:

"assignment_threshold" (float between 0 and 1): The similarity threshold

below which a landmark vector will be marked unassigned.
"good_site_normed_threshold" (float between 0 and 1): The minimum for

the cosine similarity between a good site’s representative unit vector and its best match landmark vector.
"good_site_projected_threshold" (positive float): The minimum inner product

between a good site’s representative unit vector and its best match landmark vector.
All other params are passed along to sitator.util.mcl.markov_clustering.

sitator.landmark.cluster.mcl.cov2corr(A)¶: covariance matrix to correlation matrix.

sitator.landmark.cluster.mcl.do_landmark_clustering(landmark_vectors, clustering_params, min_samples, verbose)¶