Multilevel Clustering via Wasserstein Means

Authors: Nhat Ho, XuanLong Nguyen, Mikhail Yurochkin, Hung Hai Bui, Viet Huynh, Dinh Phung

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, experiment results with both synthetic and real data are presented to demonstrate the flexibility and scalability of the proposed approach.
Researcher Affiliation Collaboration 1Department of Statistics, University of Michigan, USA. 2Adobe Research. 3Center for Pattern Recognition and Data Analytics (PRa DA), Deakin University, Australia.
Pseudocode Yes Algorithm 1 Multilevel Wasserstein Means (MWM)
Open Source Code Yes 1Code is available at https://github.com/ moonfolk/Multilevel-Wasserstein-Means
Open Datasets Yes Label Me dataset consists of 2, 688 annotated images which are classified into 8 scene categories including tall buildings, inside city, street, highway, coast, open country, mountain, and forest (Oliva and Torralba, 2001).
Dataset Splits No The paper does not provide specific details on training, validation, or testing splits for the datasets used in experiments.
Hardware Specification Yes This implementation has the advantage of making use of all of 16 cores on the test machine.
Software Dependencies No The paper mentions 'Spark-based implementation' but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes Unless otherwise specified, we set the number of groups m = 50, number of observations per group nj = 50 in d = 10 dimensions, number of global clusters M = 5 with 6 atoms. For Algorithm 1 (MWM) local measures Gj have 5 atoms each; for Algorithm 2 (MWMS) number of atoms in constraint set SK is 50.