Theoretical Analysis For Exact Results in Stochastic Linear Learning

Rezaul Karim

AbstractThis master thesis deals with the learning theory. It contains some analysis as well as derivations in Stochastic Linear learning. Derived results, for example, Generalization Error are exact in contrast with many available assumed or asymptotic results. The works are done chiefly basing on two papers: Hansen 1993 [13] and Hansen 2004 [20]. Some MATLAB simulations are done in order to prove the undoubted validity of the expressions for the Exact Generalization errors. Two expressions for the Generalization errors with respect to the sample size were derived in two different ways from linear models. The cross point of them were also detected. The properties of the curves were discussed throughout the whole sample size domain. At last, in the research part, there are some investigations about the undesired events of the curves while a discussion about the recovery from that situation is presented.
KeywordsLearning, stochastic, linear, asymptote, domain.
TypeMaster's thesis [Academic thesis]
PublisherInformatics and Mathematical Modelling, Technical University of Denmark, DTU
AddressRichard Petersens Plads, Building 321, DK-2800 Kgs. Lyngby
NoteSupervised by Prof. Lars Kai Hansen
Electronic version(s)[pdf]
BibTeX data [bibtex]
IMM Group(s)Intelligent Signal Processing

Back  ::  IMM Publications