TT8001 - Statistical Pattern Recognition
Lessons are not given in the academic year 2014/2015
The course is lectured biannually, next time spring 2016.
The course deals with statistical methods for classification and clustering.
Within classification focus is set on Bayesian theory, parametric and non-parametric techniques, different estimation methods, distortion measures, linear and nonlinear classifiers, different classifier structures, static and dynamic problems, generalization, etc.
Within clustering focus is set upon hierarchical methods, classical algorithms like K-means, newer techniques like fuzzy and competitive methods, latent semantic analysis, etc. Further, choice of distortion measures and optimization criteria matched to input room topology.
Further principles for anomaly detection will be presented.
Learning objectives : The student shall learn the theory of statistical methods for classification and clustering. This applies to both basis theory and state-of-the-art methods.
Skills : The student shall learn how to apply the theory on different physical signals like images, speech, medical signals aso.
Learning methods and activities
A combination of lectures and self-study.
Recommended previous knowledge
Knowledge of basic statistics, estimation theory and vector algebra
Required previous knowledge
Knowledge minimum comparable to course TMA4245 Statistics.
Lecture notes and journal articles/papers. The book "Pattern Recognition" by Theodoridis&Koutroumbas (AP 2006) is recommended as reference literature.