 Browse by Category

Browse by School/
Graduate School
Pattern Recognition and Exercises
Kenichiro ISHII Professor
Department: School of Engineering / Graduate School of Engineering
Class Time:  2011 Spring Tuesday 
Recommended for:  School of Infomatics and Science 
Course Overview
Course Aims
Pattern recognition is a technique for recognizing speech, images, characters and so on by computers. In this course, basic ideas of pattern recognition, classification theory, learning theory and their algorithms are introduced. A variety of exercises are presented, in order that students gain deeper understanding and acquire the ability to apply the technology to actual problems.
Key Features
 In order to provide a better understanding of the lecture, two kinds of exercises are given. One is to solve pattern recognition problems by hand, and the other is to solve problems using a computer. These problems are given in the first (lecture time) and the second (exercise time) hours, respectively.
 In the exercises, handprinted characters written by students during the course are used, so that students comprehend the technology as completely as possible.
 During each lecture some application examples and related topics are introduced by audiovisual demonstration.
 After several sessions have been completed, the course is assessed by students using questionnaires. The result of the assessment is utilized to improve the rest of the sessions.
Close Section
Syllabus
Course Objectives
The major objectives of this course are, to understand the basic ideas of pattern recognition, and to acquire skills in solving actual problems using classification/learning algorithms.
Course Outline
This course consists of two parts. The first part is a lecture, where not only technological explanation but also some exercise problems are given. The second part is an exercise for solving pattern recognition problems using a computer.
Requirements and Recommended Courses
A sufficient knowledge of linear algebra, probability theory and statistics is required. Having some programming skills is preferable in order to perform computer simulations in exercises.
Course Schedule
Session  Contents 

1  pattern recognition system, feature extraction, feature vector 
2  prototype, nearest neighbor rule, linear discriminant function 
3  perceptron learning rule, weight space, solution region 
4  perceptron convergence theorem, dimension size, sample size 
5  majority voting, piecewise linear discriminant function 
6  WidrowHoff learning rule, multiple regression analysis 
7  error estimation and perceptron 
8  back propagation method, neural network, feature evaluation 
9  transformation of feature space, Fisher's method 
10  KL expansion 
11  empirical probability, subjective probability, Bayes theorem 
12  Bayesian updating, Bayesian estimation 
13  Bayes decision rule, Bayes error 
14  parameter estimation by maximum likelihood method 
15  examination 
Grading
Grading is based on reports and the final examination.
Close Section
Page last updated March 8, 2011
The class contents were most recently updated on the date indicated. Please be aware that there may be some changes between the most recent year and the current page.