CS 801F-Pattern Recognition |
Introduction |
Examples; The nature of statistical pattern recognition; Three learning paradigms;
The sub-problems of pattern recognition; The basic structure of a pattern recognition
system; Comparing classifiers.
|
Bayes Decision Theory |
General framework; Optimal decisions; Classification; Simple performance bounds. |
Learning - Parametric Approaches |
Basic statistical issues; Sources of classification error; Bias and variance; Three
approaches to classification: density estimation, regression and discriminant
analysis; Empirical error criteria; Optimization methods; Failure of MLE;
|
Parametric Discriminant Functions |
Linear and quadratic discriminants; Shrinkage; Logistic classification; Generalized
linear classifiers; Perceptrons; Maximum Margin; Error Correcting Codes;
|
Error Assessment |
Sample error and true error; Error rate estimation; Confidence intervals; Resampling
methods; Regularization; Model selection; Minimum description length; Comparing
classifiers
|
Nonparametric Classification |
Histograms rules; Nearest neighbor methods; Kernel approaches; Local polynomial
fitting; Flexible metrics; Automatic kernels methods |
Feature Extraction |
Optimal features; Optimal linear transformations; Linear and nonlinear principal
components; Feature subset selection; Feature Extraction and classification stages,
Unsupervised learning and clustering, Syntactic pattern recognition, Fuzzy set
Theoretic approach to PR,
|
Margins and Kernel Based Algorithms |
Advanced algorithms based on the notions of margins and kernels
|
Applications of PR |
Speech and speaker recognition, Character recognition, Scene analysis.
|