Computational Intelligence Society DL in Tunisia
Co-sponsored by National School of Engineers of Sfax (ENIS) and Research Group on Intelligent Machines (REGIM)
Seminars by Prof. Jim Bezdek from Univ. of West Florida, USA
|
Location: Building: Vincci Resort Hotel 4* Room Number: Seminar Room Djerba, Tunisia |
|
|
Start time: 05-Nov-2009 03:00PM End time: 06-Nov-2009 06:00PM All times are: Africa/Tunis |
|
| IEEE CIS Tunisia Chapter | |
iCalendar (2.0-ical): Download or import iCal file into Calendar program (e.g. Outlook)
Admission Charge applies.
Survey: Fill out the survey
Speakers:
Jim Bezdek of University of West Florida
Topic: Visual Clustering in Relational Data
This talk begins by defining the three canonical problems of clustering (assessment, clustering, and validation). Then I will give a brief history of visual approaches to these three problems. Following this, I will discuss a little theory and some applications of (8) visual algorithms. I will explain how each of these algorithms operates, and illustrate various facets of each with some simple examples.
Email:jbezdek@uwf.edu
Address: Florida, United States, 32514
Topic: Visual Clustering in Relational Data
This talk begins by defining the three canonical problems of clustering (assessment, clustering, and validation). Then I will give a brief history of visual approaches to these three problems. Following this, I will discuss a little theory and some applications of (8) visual algorithms. I will explain how each of these algorithms operates, and illustrate various facets of each with some simple examples.
Email:jbezdek@uwf.edu
Address: Florida, United States, 32514
Jim Bezdek of University of West Florida
Topic: Fuzzy clustering in very large data sets
This talk focuses on adaptations of the fuzzy c-means, expectation-maximization, and non-Euclidean relational c-means clustering algorithms for use with very large data sets. I will discuss three cases, according as the data are image data, feature vector data, or square relational data. There are two general approaches to such problems: distributed clustering on many processors; or sampling and extension. The methods I discuss are based on sampling+extension. I will show that sampling needs to be different for each case, and extension follows naturally for many clustering algorithms from first order necessary conditions for the models being generalized. Numerical examples illustrate each approach. The results I present in this talk are ok, but there is plenty of room for improvement.
Email:jbezdek@uwf.edu
Address: Florida, United States, 32514
Topic: Fuzzy clustering in very large data sets
This talk focuses on adaptations of the fuzzy c-means, expectation-maximization, and non-Euclidean relational c-means clustering algorithms for use with very large data sets. I will discuss three cases, according as the data are image data, feature vector data, or square relational data. There are two general approaches to such problems: distributed clustering on many processors; or sampling and extension. The methods I discuss are based on sampling+extension. I will show that sampling needs to be different for each case, and extension follows naturally for many clustering algorithms from first order necessary conditions for the models being generalized. Numerical examples illustrate each approach. The results I present in this talk are ok, but there is plenty of room for improvement.
Email:jbezdek@uwf.edu
Address: Florida, United States, 32514
