Frugal and Online Affinity Propagation

Xiangliang Zhang 1 Cyril Furtlehner 1 Michèle Sebag 1
1 TAO - Machine Learning and Optimisation
CNRS - Centre National de la Recherche Scientifique : UMR8623, Inria Saclay - Ile de France, UP11 - Université Paris-Sud - Paris 11, LRI - Laboratoire de Recherche en Informatique
Abstract : A new Data Clustering algorithm, Affinity Propagation suffers from its quadratic complexity in function of the number of data items. Several extensions of Affinity Propagation were proposed aiming at online clustering in the data stream framework. Firstly, the case of multiply defined items, or weighted items is handled using Weighted Affinity Propagation(WAP). Secondly, Hierarchical AP achieves distributed AP and uses WAP to merge the sets of exemplars learned from subsets. Based on these two building blocks, the third algorithm performs Incremental Affinity Propagation on data streams. The paper validates the two algorithms both on benchmark and on real-world datasets. The experimental results show that the proposed approaches perform better than $K$-centers based approaches.
Document type :
Conference papers
Complete list of metadatas

Cited literature [14 references]  Display  Hide  Download

https://hal.inria.fr/inria-00287381
Contributor : Xiangliang Zhang <>
Submitted on : Wednesday, June 11, 2008 - 5:15:58 PM
Last modification on : Thursday, April 5, 2018 - 12:30:12 PM
Long-term archiving on : Friday, September 28, 2012 - 3:51:21 PM

File

v5_for_final.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00287381, version 1

Collections

Citation

Xiangliang Zhang, Cyril Furtlehner, Michèle Sebag. Frugal and Online Affinity Propagation. Conférence francophone sur l'Apprentissage (CAP), May 2008, Ile de Porquerolles, France. ⟨inria-00287381⟩

Share

Metrics

Record views

540

Files downloads

379