# CSCE 768 Pattern Recognition and Classification

Prerequisites: STAT 509 (Statistics for Engineers) or 510 (Introduction to Applied Probability) or 511 (Probability)

Meeting Time and Place: TTh 930-1045, Sumwalt 241.

Instructor: Marco Valtorta
Office: Sumwalt 206A, 777-4879
E-mail: mgv@cs.sc.edu
Office Hours: Mon 3-5, Wed 3-5,
Grader: No grader for this course

## Goals of the Course

The goals of this course, as adapted from the old departmental syllabus, are:

• To introduce the field of pattern recognition, including feature extraction, pattern classification, and cluster analysis.
• To describe the main statistical pattern recognition approaches and algorithms.
• To present statistical decision theory.
• To describe linear and nonlinear classifiers, including neural networks and their relation to Bayesian decision theory.
• To study the foundations of the theory of classifiers.
• To study issues of sample and time complexity in pattern classification or concept learning.

## Grading Policy

The precise grading policy has not yet been determined. The course grade will be based on homework, a midterm exam (or possibly two midterm exams), a final exam, possible quizzes. The homework consists of both paper and pencil exercises and computer exercises using Maple. A final project involves the analysis of a data set using a classifier such as Parzen windows, the k-nearest-neighbors method, etc. While the final project can be carried out using Maple, most students in the past found it more convenient to implement the algorithm using a traditional imperative programming language such as C.

## Code of Student Academic Responsibility

You are expected to be aware of and to follow the academic code of responsibility that appears in the Carolina Community Student Policy Manual. Except for explicitly designated team assignments, all work that is turned in is expected to be your own.

Here is the reference to another text on statistical pattern recognition: Fukunaga, Keinosure. Introduction to Statistical Pattern Recognition (second edition). Academic Press, 1990. ISBN: 0-12-269851-7.

Some useful links:
A talk on hidden Markov models that emphasizes the connection to Bayesian nets (in postscript).
How to view postscript in Windows: Wim Sweldens's web page on GSview.
A handout with Maple examples related to Exercise 14, Ch.2 DHS01.
An introduction to Maple: Professors Miller and Meade's "Day 1" lecture.
A Talk by Kathy Laskey: Bayesian Decision Theory and Machine Learning.