o:1009 Treniranje strukturnih klasifikatora za različite funkcije gubitaka sa primenom na probleme klasifikovanja sekvenci : doktorska disertacija en This thesis presents algorithms for training structured classifiers over different loss functions. It introduces a new primal subgradient method for the optimization of average sum loss, several extensions of max-margin classifiers to the k-best case, a sequential dual method for the structured ramp loss optimization, It considers different decoding algorithms over semirings and presents an organization of a two-structured-model committee in order to improve the results. It includes theoretical analysis of introduced algorithms, as well as experimental results on sequence labelling problems in natural language processing. Umnoženo za odbranu. Univerzitet u Nišu, Prirodno-matematički fakultet, 2015. Bibliografija: str. 87-93. Izvod ; Abstract. artificial intelligence Datum odbrane: 06.07.2015. 1025057769 91552101 2252 91552100 1025057769 2016-01-22T15:30:32.053Z 45 no 46 mentor Dejan, 1985- Mančev 2015 63 mentor Branimir 1967- Todorović 2015 63 član komisije Miroslav 1964- Ćirić 2015 63 član komisije Predrag 1959- Stanimirović 2015 63 član komisije Miomir Stanković 2015 63 član komisije Leonid Stoimenov 2015 PDF/A (144 lista) 4030838 http://phaidrabg.bg.ac.rs/o:1009 no yes 4 2016-01-22T15:30:32.320Z 70 Računarske nauke, strukturno učenje, klasifikacija sekvenci structured learning, sequence labeling 004.8 1738 18A07 18A0701 2015