The PDM package carries out the Partition Decoupling Method (PDM), an unsupervised machine-learning technique that consists of two iterated submethods: a spectral clustering step that finds the dominant clusters of samples, and a "scrubbing" step that removes this structure (by projecting the data onto the cluster centroids and taking the residuals) such that the next clustering iteration can articulate an independent set of clustering relationships. The method is fully described in:
If you use the PDM package or source code, please cite:
Please email Rosemary Braun at rosemary.braun_AT_GMAIL_DOT_COM if you have any questions, concerns, or feedback!