Course Outline

3 credits, lecture
Semester taught: Fall
Prerequisite: Stat 217 or IME 354 (knowledge of statistics)

Statistical, syntactic, neural and fuzzy set approaches to pattern recognition. Pattern recognition is a fundamental science for artificial intelligence, robotics, information retrieval systems and other computer applications. This course emphasises the understanding of basic concepts and uses example data sets to improve intuitive understanding.

Course Goals:

• Understand the place of pattern recognition in machine learning.
• Be familiar with the topology of pattern recognition techniques.
• Understand the pattern recognition design cycle the relationship between data, features, models, training and evaluation.
• Be capable of applying the principles of Bayesian Decision Theory to pattern recognition problems and the distinctions between parametric and non-parametric methods.
• Be capable of applying linear discriminant function analysis to pattern recognition problems.
• Understand the fundamentals of neural networks and be capable of using Backpropogation and Radial Basis Functions.
• Be familiar with algorithm-independent machine learning and unsupervised learning.
• Be capable of using some of the basic software tools to solve pattern recognition problems.

Syllabus:

Note: The syllabus is informational, and may not have any relationship to the actual sequence of events in the course.


• A review of important statistical principles
• An introduction to pattern recognition concepts
• Statistical pattern recognition and Bayesian classification
• Linear discriminants
• Supervised learning methods
• Unsupervised learning
• Syntactic pattern recognition and learning grammers
• Neural network principles
• Neural associative mapping
• Neural feedforward and backpropagation methods
• Content addressable memory neural networks
• Fuzzy logic and applications to pattern recognition
  

Fall 2004 Details

Lecture: MWF, 8:00 - 8:50 , EPS 108 (Subject to change)

The course will study the concepts of statistical pattern recognition in some detail, as that is the foundation for all other methods. This study will be organized around a series of problems and example data sets. Following this, syntactic and neural network methods of pattern recognition will be studied and related to statistical methods. Time permitting, fuzzy set methods will be investigated.

• Pattern Classification, 2nd Ed., Duda, Hart & Stork, Wiley
  
• Computer Manual in Matlab, Stork & Yom-Tov, Wiley (Optional)