Course Name Cluster Analysis
Instructor and position Michael Houle, Visiting Professor
Students 1-5th year
Credits 2
Lecture/experiment Lectures
Summary This course deals with the theoretical and practical issues surrounding the topic of cluster analysis for knowledge discovery. A comparative review of clustering strategies will be presented, as well as their applications, and the data structures needed to support them. Particular attention will be given to the implications of data representations and algorithmic design choices on the scalability and applicability of the various approaches studied.
Aims An understanding of the main issues surrounding cluster analysis, particularly as regards implementation. Familiarity with state-of-the-art methods. The ability to assess the scalability and effectiveness of clustering strategies.
Plan

(1) Clustering versus classification. 
(2) Features, similarity and search.
(3) Scalability and the course of dimensionality.
(4) Overview of clustering strategies.
(5) Partitional clustering. 
(6) Hierarchical methods. 
(7) Grid-based methods. 
(8) Model-based methods.
(9) Density-based methods. 
(10) Shared-neighbor methods.
(11) Other clustering strategies.
(12) Application: outlier detection.
(13) Application: feature selection.
(14) Application: query result clustering.
(15) Other applications.
Evaluation criteria Assessment based on class participation, reports, and presentations.
Textbooks None. 
References Research papers.
Prerequisites Basic knowledge of artificial intelligence and computer science.
Remarks None.

 

-close-