Home
News
Call for papers
Important dates
Committees
Special sessions
Tutorials
Demo sessions
Invited speakers
Paper submission
Technical program
Venue
Sponsors
Contact
Registration
Important Deadlines
Regular sessions extended paper submissions :
September 1st 15th, 2008


Special Sessions extended paper submission :
September 1st 15th, 2008


Demo Sessions submission :
September 1st 15th, 2008


Special Sessions Authors / Demo notification :
October 12th, 2008


Regular sessions Author notification :
October 12th, 2008


Final Paper Submission and Registration :
November 1st, 2008


    

TITLE: Clustering: Algorithms and Applications.

AUTHOR: Hichem Frigui, Associate Professor and Director of Multimedia Research Lab Department of Computer Engineering and Computer Science, University of Louisville (USA)

ABSTRACT:

Unsupervised learning, or clustering, is an effective technique for exploratory data analysis, and has been studied extensively in statistics, pattern recognition, machine learning, data mining, image analysis, and multimedia information retrieval. This talk will have two main parts. The first one will focus on outlining several approaches to clustering. Both crisp and fuzzy clustering methods will be discussed, and we will outline possible solutions to the following main research issues in clustering:

  1. Scalability to the size of the data;
  2. Ability to handle arbitrary, domain-specific subjective dissimilarity measures;
  3. Ability to identify clusters that are dense in only subspaces of the original high-dimensional data space;
  4. Robustness in the face of noisy data; and
  5. Ability to automatically determine the number of clusters. The second part of my talk will focus on various applications of clustering algorithms. This includes :
  • (i) content-based image retrieval (CBIR);
  • (ii) image database categorization and visualization;
  • (iii) video summarization;
  • (iv) handwritten word recognition;
  • (v) image segmentation; and
  • (vi) text mining and information retrieval.

We will focus on the categorization of a large collection of images. We will show that this organization can be used to build an efficient indexing structure, build an adaptive navigation system, show the user the most representative images in a query by visual example, and develop an adaptive weighted similarity measure where the weights are category dependent. For each of the above applications, we define the problem, illustrate the need for clustering to solve the problem, and show some results.

Invited speakers



Powred by : MousePlus
Copyright 2008 Tunisien Scientific Society. All rights reserved
Call for papers Important dates Paper submission