Theme
This lecture has been given in October 2010 at the
Computational Intelligence and Learning
doctoral school and May 2012 at the University of Málaga
(at Faculty of Telecommunication Engineering).
It is a three to five hours introduction to deterministic
annealing as a general framework for combinatorial optimization. The course is
self contained, but some background in optimization and probabilities is
needed.
Slides
The slides are available here.
References
The course is based on my research work and on the excellent survey about
deterministic annealing written by Kenneth Rose. Papers available online include:
- Kenneth Rose survey:
Deterministic annealing for clustering, compression, classification, regression, and related optimization problems, Proceedings of the IEEE, 86 (11):
2210–2239, November 1998
- Dissimilarity clustering:
Pairwise data clustering by deterministic annealing, Thomas Hofmann & Joachim
Buhmann. IEEE Transactions on Pattern Analysis and Machine Intelligence,
1997, Vol. 19(1), pp. 1-14 (and 19(2), p.192)
- Graph clustering for visualization: Optimizing an Organized Modularity Measure for Topographic Graph Clustering: a Deterministic Annealing Approach,
Fabrice Rossi and Nathalie
Villa-Vialaneix. Neurocomputing, volume 73, number 7-9, pages
1142-1163. March 2010.