ITL Class


Graduate 
Prereq: EEL 6502 or approval of instructor. 
Catalog 
Fundamentals of adaptive systems training and information theory. Renyi’s entropy definition. Sample estimators of Renyi’s entropy. Cost functions based on entropy and divergence and their corresponding training algorithms. Reproducing Kernel Hilbert spaces (RKHS) and ITL. Correntropy as a new similarity measure. Statistical signal processing in RKHS. Applications in signal processing and machine learning 
Text Book 
Instructor notes (being compiled in a book). 
References 
Haykin S., Adaptive Filter Theory, Prentice Hall (4th edition) 
Professor 
Jose C. Principe, Professor of Electrical Engineering. Office EB 451, email: principe@cnel.ufl.edu 
Goals 
Understand the principles of inference to implement nonlinear signal processing algorithms. New cost functions to train adaptive signal processing algorithms will be presented and compared with conventional adaptive filters. Links between ITL and reproducing kernel Hilbert spaces will be established. New RKHS will be presented that are specifically useful for statistical signal processing and machine learning. 
Grading 
Homework........20% 
Computer 
Homework and projects will require access to a fast personal computer to MATLAB or equivalent. Familiarity with Matlab or another simulation language is essential 
Follow these Guidelines to edit the ITL Wiki.