"1"^^ . . . "29"^^ . . "2008-06-02+02:00"^^ . . "Within the framework of the research project we concentrated on theoretical development of a specific class of models designed for uncertain knowledge representation and processing: models based on both probability and possibility multidimensional distri"@en . "Neuvedeno."@en . "29"^^ . . "0"^^ . . . . . "GA201/02/1269" . "Po celou dobu trv\u00E1n\u00ED projektu byla hlavn\u00ED pozornost soust\u0159ed\u011Bna na teoretick\u00FD rozvoj apar\u00E1tu pro modelov\u00E1n\u00ED nejist\u00FDch znalost\u00ED pomoc\u00ED mnohorozm\u011Brn\u00FDch pravd\u011Bpodobnostn\u00EDch i posibilistick\u00FDch distribuc\u00ED. Studovali jsme klasick\u00E9 bayesovsk\u00E9 s\u00EDt\u011B i modely zalo"@cs . . . "http://www.isvav.cz/projectDetail.do?rowId=GA201/02/1269"^^ . . "Mnohorozm\u011Brn\u00E9 strukturovan\u00E9 modely" . . . . . "Multidimensional structured models"@en . "Graphical Markov models, one of the most popular tools of modern multivariate statistics used for description of multidimensional structured models, is the very technique that dominates probabilistic modelling in Artificial Intelligence. On the other hand, in the last two decades possibility theory has emerged and its applicability in AI has been proven as an alternative to probability theory. Furthermore, within possibilistic framework, several attempts have been made to construct models corresponding to some graphical Markov models, namely Bayesian networks, and properties of conditional possibilistic independence suggest that alternatives of a wider class of graphical models can be defined. Therefore, we propose to investigate and develop new parallels in these two fields, to distil their common features and, simultaneously, not to confine ourselves to graphical models. Our theoretical goal is to obtain a unified view on both kinds of models for which we will employ the apparatus"@en . . . "0"^^ . . "Grafick\u00E9 markovsk\u00E9 modely, jeden z nejroz\u0161\u00ED\u0159en\u011Bj\u0161\u00EDch n\u00E1stroj\u016F modern\u00ED statistiky pro popis mnohorozm\u011Brn\u00FDch strukturovan\u00FDch model\u016F, jsou technikou, kter\u00E1 dominuje pravd\u011Bpodobnostn\u00EDmu modelov\u00E1n\u00ED v um\u011Bl\u00E9 inteligenci. Na druhou stranu, v pr\u016Fb\u011Bhu uplynul\u00FDch dvaceti let se objevila teorie mo\u017Enosti (possibility theory) a prok\u00E1zala svoji pou\u017Eitelnost v oblasti um\u011Bl\u00E9 inteligence jako alternativa k teorii pravd\u011Bpodobnosti. Nav\u00EDc se v r\u00E1mci teorie mo\u017Enosti objevilo n\u011Bkolik pokus\u016F konstruovat modely odpov\u00EDdaj\u00EDc\u00ED n\u011Bkter\u00FDm grafick\u00FDm markovsk\u00FDm model\u016Fm, konkr\u00E9tn\u011B bayesovsk\u00FDm s\u00EDt\u00EDm, a vlastnosti podm\u00EDn\u011Bn\u00E9 posibilistick\u00E9 nez\u00E1vislosti nazna\u010Duj\u00ED, \u017Ee bude mo\u017En\u00E9 definovat prot\u011Bj\u0161ky \u0161ir\u0161\u00ED t\u0159\u00EDdy grafick\u00FDch model\u016F. Proto navrhujeme zkoumat a rozv\u00EDjet nov\u00E9 paralely mezi t\u011Bmito dv\u011Bma oblastmi, odhalovat jejich spole\u010Dn\u00E9 rysy, a p\u0159itom se neomezovat jen na grafick\u00E9 modely. Teoretick\u00FDm c\u00EDlem je z\u00EDskat jednot\u00EDc\u00ED pohled na oba druhy model\u016F; zde hodl\u00E1me vyu\u017E\u00EDt apar\u00E1t zalo\u017Een\u00FD na iterativn\u00EDm pou\u017Eit\u00ED oper\u00E1tor\u016F skl\u00E1d\u00E1n\u00ED. Tuto" . . .