Computational methods for the identification and classification of objects. Feature extraction, feature-space representation, distance and similarity measures, decision rules. Supervised and unsupervised learning. Statistical pattern recognition: multivariate random variables; Bayes and minimum-risk decision theory; probability or error; feature reduction and principal components analysis; parametric and nonparametric methods; clustering; hierarchical systems. Syntactic pattern recognition: review of automata and language theory; shape descriptors; syntactic recognition systems; grammatical inference and learning. Artificial neural networks as recognition systems.
Why take this course?
Pattern recognition is important in many fields related to electrical and computer engineering, including signal analysis, image analysis, and communication theory.
The course assumes knowledge of probability and random variables, as introduced in STAT 4714.
Major Measurable Learning Objectives
design and implement algorithms that can perform pattern recognition
develop problem-specific similarity measures;
compute the probability of classification error when underlying probability distributions are known.
Percentage of Course
Review of Statistical Methods
Linear and Piecewise-Linear Discriminate Design
Review of Automata Theory and Formal Languages
Grammatical Inference; Learning in Syntactic Recognition