Last update: RNDr. Michal Kopecký, Ph.D. (09.05.2019)
With the current flood of information and services on the Web it is necessary to have models of information processing which provide ordering of results by relevance tailored to each user/customer separately.
The aim of the lecture is to inter-link several information models (mainly formal declarative / deductive and inductive) and extend them with ordering.
Solution is demonstrated by LMPM – Linear Monotone Preference Model. In labs we work with some formal models (preferential Datalog), graphical version of LMPM and experiments on small illustrative data.
Last update: RNDr. Filip Zavoral, Ph.D. (26.04.2021)
Předmět je zrušen, látka je obsažena v NSWI166, který s tímto předmětem záměnný.
Course completion requirements -
Last update: prof. RNDr. Peter Vojtáš, DrSc. (27.10.2020)
Following conditions are "corona"-dependent subject to change, actual form of deliverables and timetable will be settled and announced on classroom page for definitely enrolled students when "corona"-conditions change.
Terms of passing the course consist of homework, mainly on paper and some about experimenting with small illustrative data. These are only conditions for getting credits.
Exam is oral and requires basic understanding of the whole material.
As soon as terminology is introduced, detailed milestones (also form of deliverables) and preferred deadlines (with possible repeated attempts) will be announced at a lab.
There is no evidence on personal presence. Nevertheless, no additional explanation for tasks will be given, except on the respective lab and brief description on the course web. Final deadline is end of semester.
Last update: prof. RNDr. Peter Vojtáš, DrSc. (27.10.2020)
Following conditions are "corona"-dependent subject to change, actual form of deliverables and timetable will be settled and announced on classroom page for definitely enrolled students when "corona"-conditions change.
Terms of passing the course consist of homework, mainly on paper and some about experimenting with small illustrative data. These are only conditions for getting credits.
Exam is oral and requires basic understanding of the whole material.
As soon as terminology is introduced, detailed milestones (also form of deliverables) and preferred deadlines (with possible repeated attempts) will be announced at a lab.
There is no evidence on personal presence. Nevertheless, no additional explanation for tasks will be given, except on the respective lab and brief description on the course web. Final deadline is end of semester.
Literature -
Last update: RNDr. Michal Kopecký, Ph.D. (09.05.2019)
Fagin, Lotem, Naor. Optimal aggregation algorithms for middleware, J. Computer and System Sciences 66 (2003), pp. 614-656, http://researcher.watson.ibm.com/researcher/files/us-fagin/jcss03.pdf
Supporting material on the course web
Last update: RNDr. Michal Kopecký, Ph.D. (09.05.2019)
Fagin, Lotem, Naor. Optimal aggregation algorithms for middleware, J. Computer and System Sciences 66 (2003), pp. 614-656, http://researcher.watson.ibm.com/researcher/files/us-fagin/jcss03.pdf
Učební materiál na webu předmětu
Syllabus -
Last update: prof. RNDr. Peter Vojtáš, DrSc. (23.05.2020)
Information models and ordering
Motivation problems, use-case, data, challenge, goal, who/what is better, ordering as preference
Various representation and presentations of ordering in data, information, knowledge
Linear Monotone Preference Model
Fagin's data model and threshold top-k algorithm - Deduction-Querying, Search, Retrieval, ...
Linear Monotone Preference Model a Fagin’s data model
Threshold algorithm, Correctness
Measures of success of preference learning algorithms - Induction-Learning, Generalization, Estimation, Prediction
Class of models
System of metrics
Experiments, validation, evaluation
Datalog / logic programming - logical/relational domain calculus with ordering
Many valued characteristics of sets as a tool for coding ordering, many valued logic, connectives
Many valued modus ponens - declarative models - model of deduction, induction, querying
Many valued modus ponens, residuated operators and correctness
Many valued logic programming and correctness
Last update: prof. RNDr. Peter Vojtáš, DrSc. (23.05.2020)
Informační modely a uspořádání
Motivační problémy, use-case, data, výzva, cíl, co/kdo je lepší, uspořádaní jako preference
Různé způsoby reprezentace a prezentace dat, informací, znalostí
Lineární Monotónní Preferenční Model
Faginův datový model a prahový top-k algoritmus - dedukce - dotazování, vyhledávaní,…
Lineární Monotónní Preferenční Model a Faginův datový model
Prahový algoritmus, korektnost
Míry úspěšnosti algoritmů učení preference - indukce - učení, zobecnění, odhad, předpověď
Třídy modelů
Systémy metrik
Experimenty v informatice, validace, vyhodnocení
Datalog/logické programování - logický/relační doménový kalkul s uspořádáním
Vícehodnotové charakteristické funkce množin jako kódovaní uspořádání, vícehodnotová logika, spojky
Vícehodnotový modus ponens - Deklarativní modely - model dedukce, indukce, dotazování
Vícehodnotový modus ponens, reziduované operátory a korektnost
Vícehodnotové logické programování - korektnost modelu
