Hierarchical Mode Association Clustering (HMAC)


Packages in C, Matlab, R are implemented for these algorithms:
7 reasons for using MAC and HMAC:
  1. MAC requires no model fitting and uses a nonparametric kernel density estimation.
  2. The density of a cluster is not restricted to be parametric, for instance, Gaussian, but ensures uni-modality.
  3. MAC is similar to mixture-model-based clustering in the sense of characterizing clusters by smooth densities.
  4. HMAC seems to combine the complementary merits of bottom-up clustering such as linkage and top-down clustering such as mixture modeling and k-means.
  5. MAC can be performed hierarchically or in a one level manner.
  6. For hierarchical clustering, HMAC can either generate a nested structure as a conventional dendrogram, or a non-nested hierarchy.
  7. Ridgelines between pairs of clusters can be computed to numerically profile the separation between clusters, a useful feature for diagnosis.

C package (by Jia Li):
Matlab and R packages (by Yeojin Chung, with documentation):
Related publication:

J. Li, S. Ray, B. G. Lindsay, "A nonparametric statistical approach to clustering via mode identification," Journal of Machine Learning Research , 8(8):1687-1723, 2007. ( download )

Reference requirement:
  1. If you use codes downloaded from this site, please reference to the JMLR paper above and this Web site.
Copyright notice:
  1. Copyright of the C codes belongs to Jia Li, that of the Matlab and R codes belongs to Yeojin Chung.
  2. This package is for research only. Commercial usage without the agreement of the copyright owner is not allowed.
  3. Do NOT redistribute the codes.
  4. For questions on the codes, please contact Jia Li (jiali at psu dot edu) and Yeojin Chung (yxc228 at psu dot edu) respectively.


@Jia Li, Updated August, 2007          Back to Home