Budapest University of Technology and Economics, Faculty of Electrical Engineering and Informatics

    Belépés
    címtáras azonosítással
    vissza a tantárgylistához   nyomtatható verzió    

    Cooperation and Machine Learning Laboratory

    A tantárgy neve magyarul / Name of the subject in Hungarian: Kooperáció és gépi tanulás labor

    Last updated: 2008. szeptember 1.

    Budapest University of Technology and Economics
    Faculty of Electrical Engineering and Informatics

     

    		Applied Informatics Branch 
    Intelligent Systems Specialization
    MSc
    Course ID Semester Assessment Credit Tantárgyfélév
    VIMIM223 2 0/0/3/f 4  
    3. Course coordinator and department dr. Strausz György,
    4. Instructors
    Name: Position: Department:
    Dr. György Strausz
    Dániel László Kovács     		associate professor assistant lecturer DMIT
    DMIT
    5. Required knowledge Artificial Intelligence

     
    6. Pre-requisites
    Ajánlott:
    None
    7. Objectives, learning outcomes and obtained knowledge The laboratory is divided into thematic blocks, which are smaller projects, where students try to reach a given objective. Knowledge modeling and information retrieval block: where student implement more complicated, intelligent information search methods in a given domain. Cooperation block: where students implement agent societies participating in electronic auctions and voting via a game theoretic approach. Planning block: where students solve a planning and scheduling task via different AI planning methods. Learning block: where students experiment with static, dynamic, and Bayesian learning schemes in a given problem domain.

     
    8. Synopsis

    I. Simple text retrieval. The goal is to try several simpler text indexing and retrieval methods.

    II. Domain modeling. The goal is to create a model of the domain necessary for the semantic search, and to familiarize with the Protégé ontology editing software tool.

    III. Semantic information retrieval. The goal is to mend the results of the previous two steps: by using the model of the domain (the ontology) students extend the simple information retrieval with semantic capabilities.

    IV. Game theoretic agents. Student experiment with several game theoretic models (games) by implementing them with simple JADE (Java Agent DEvelopment framework) agents. Cooperative, non-cooperative behaviors and equilibrium situations are examined.

    V. Auctions and voting protocols. By using the standard message and protocol schemas, students build simple agent societies from simple JADE agents, and implement and manipulate more sophisticated auctions and voting protocols.

    VI. Single-agent (centralized) planning. Students need to represent realistic planning domains and problems with an appropriate planning problem description language. The finished planning problem/domain representation is given as an input to a “black box” planner, which automatically computes the plan(s) solving the problem.

    VII. Multi-agent (decentralized) planning. Students familiarize themselves with really distributed, multi-agent planning. The task of the students is to implement autonomous planning agents with BDI (Belief-Desire-Intention) architecture, which realize PRS-like (Procedural Reasoning System) reactive planning.

    VIII. Static neural networks. Students construct several types of static neural networks to test the effect of different parameter settings in case of simpler classification tasks.

    IX. Predicting time-series with dynamic networks. Students construct a system able to effectively predict the following element, or tens of elements in a ready-made data series by using dynamic networks (MLP, RBF, or SVM).

    X. Bayesian learning. The goal is to examine domain model learning based on passive observations via Bayesian networks.

    Theoretical knowledge related to the subject is presented within the Cooperation and Intelligence.

    9. Method of instruction Laboratory. Students participate in 10 laboratory (4 hours each) practices in a row.
    10. Assessment

    Finishing every practice and deliver every associated report (the guidelines for making these reports are specified in the practice instructions accordingly).

    The final mark is based on the different report marks.

    11. Recaps Failed or absent practices can be repeated right after the last practice in the midterm. Maximally 2 practices can be repeated.

     
    12. Consultations As required, upon agreement.
    13. References, textbooks and resources

    Stuart Russell and Peter Norvig: Artificial intelligence. a modern approach, 2nd edition, Prentice Hall, 2003

    Notes for the laboratory work available on the web page of the subject (under preparation)
    14. Required learning hours and assignment
    Kontakt óra42
    Félévközi készülés órákra30
    Felkészülés zárthelyire
    Házi feladat elkészítése
    Kijelölt írásos tananyag elsajátítása48
    Vizsgafelkészülés
    Összesen120
    15. Syllabus prepared by
    Name: Position: Department:
    Dr. György Strausz
    Dániel László Kovács     		associate professor assistant lecturer DMIT
    DMIT