Unsupervised and semi-supervised clustering by message passing: Soft-constraint affinity propagation

TL;DR

SCAP is a novel clustering method inspired by statistical physics that efficiently uncovers hierarchical structures and can incorporate partial labels for semi-supervised learning, showing promising results on artificial and biological data.

Contribution

Introduces SCAP, a new message-passing clustering algorithm that is both efficient and adaptable to semi-supervised scenarios, with detailed performance analysis.

Findings

Efficiently uncovers hierarchical data structures.

Effectively incorporates partial labels in clustering.

Performs well on artificial and biological datasets.

Abstract

Soft-constraint affinity propagation (SCAP) is a new statistical-physics based clustering technique. First we give the derivation of a simplified version of the algorithm and discuss possibilities of time- and memory-efficient implementations. Later we give a detailed analysis of the performance of SCAP on artificial data, showing that the algorithm efficiently unveils clustered and hierarchical data structures. We generalize the algorithm to the problem of semi-supervised clustering, where data are already partially labeled, and clustering assigns labels to previously unlabeled points. SCAP uses both the geometrical organization of the data and the available labels assigned to few points in a computationally efficient way, as is shown on artificial and biological benchmark data.