Proposed Course Outline: Pattern Recognition
1. Introduction - What is Pattern Recognition
Interdisciplinary Nature: Who uses Pattern Recognition
?
Mathematics and Statistics, Image analysis, Speech
Processing, Artificial Intelligence, even Psychology !
Brief Outline of Three Principal Paradigms
Statistical Pattern Recognition - oldest,
most important, covered in detail
Neural Pattern Recognition
Syntactic Pattern Recognition - touched
briefly in a separate module
`Flavour' of Prequisites
Relevant topics in
2. What does my data look like: Do I need
all of it ?
Introducing Statistical Analysis, Principal Component
Analysis
Feature Vector Representation, Normalization
Eigenvector-Eigenvalue Transformation (KLT)
Singular Value Decomposition (SVD) and Applications
3. Important Paradigm: Drawing the Line
Getting a Decision Boundary
Comparing two patterns, and Distance Measures
Comparing two patterns
-
Isolated Patterns
-
Patterns from a group of patterns
Distance Measures
Minkowski, Euclidean, Mahalanobis
Bayesian Decision Theory
Classifiers, Discriminant Functions, Decision Surfaces,
Maximum Likelihood Estimation (MLE), Expectation Maximization (EM)
4. Neural Networks: A Brief Introduction
Small Module
5. Main Application: Unsupervised Learning and
Clustering
5a. Statistical Approaches
-
Bayes Decision Theory (Covered
in detail above)
-
K-Means Algorithm
-
Neirest Neighbour Classification
-
Hierarchical Methods
-
Graph-Theoretic Methods
-
Simulated Annealing
5b. Neural Approaches
-
Kohonen's Net, and Self-Organizing Maps (Neural
Equivalent of K-Means)
-
Adaptive Resonance Theory (ART)
-
Other Similar Approaches for Neural Networks
6. Syntactic Pattern Recognition
A Brief Introduction
7. Miscellaneous Topics
(N)one or more, as time permits
Hidden Markov Models (HMMs)
Support Vector Machines (SVMs)
Sumantra Dutta Roy
Department of Electrical Engineering, IIT Bombay, Powai,
Mumbai - 400 076, INDIA. sumantra@ee.iitb.ac.in