Syllabus

Csc. 6724 : Pattern Recognition Fall 2006

Professor : Octavio Betancourt Office : NAC 8/216A

Phone : 650-6180

Main Text : " Pattern Recognition", S. Theodoridis and K. Koutroumbas, Academic Press, 1999.
Other Texts : " Neural Networks : A comprehensive Foundation ", Simon Haykin. Macmillan, New York. 1994

" Neural Networks for Pattern Recognition ", Christopher M. Bishop, Oxford University Press, 1995.

" Neural Networks and Intellect ", Leonid Perlovsky, Oxford University Press, 2001.

Pre - Requisite : Csc. 301, Csc. 217.

Grading : 3 Projects :

1.- Bayesian approach to the classification problem for hand written digits

( 25 % ).

2.- Multi-layered, feed forward network training , parity function calculation ( 25 % )

3.- Application ( Optical Character recognition or other ) ( 50 % )

Material to be Covered

I.- Introduction

1.- Is Pattern Recognition Important ?

2.- Features, Feature Vectors, and Classifiers

3.- Supervises versus Unsupervised Pattern Recognition

II.- Classifiers based on Bayes Decision Theory

1.- Introduction

2.- Bayes Decision Theory

3.- Discriminant Functions and Decision Surfaces

4.- Bayesian Classification for Normal Distributions

5.- Estimation of Unknown Probability Density FunctionsT)

6.- The Nearest Neighbor Rule

III.- Linear Classifiers

1.- Introduction

2.- Linear Discriminant Functions and Decision Hyperplanes

3.- The Perceptron Algorithm

4.- Least Squares Methods

IV.- Nonlinear Classifiers

1.- Introduction

2.- The XOR Problem

3.- The Two-layer Perceptron

4.- Three-Layer Perceptrons
5.- The Backpropagation Algorithm

6.- Choice of Cost Function, Network size, weight sharing

7.- Radial Basis Function Networks

V.- Feature Selection

1.- Introduction

2.- Preprocessing

3.- Statistical Hypothesis Testing

4.- Class Separability Measures

5.- Feature subset selection

VI.- Feature Generation

1.- Introduction

2.- Basis Vectors and Images

3.- The Karhunen-Loeve Transform

4.- The Discrete Fourier Transform

5.- The Discrete Cosine and Sine Transforms

6.- Features for Shape and Size Characterization

VII .- Netlab, a general software package for Pattern Recognition