Vai al contenuto principale
Oggetto:

Probabilistic Graphical Models in Intelligent Systems

Oggetto:

Probabilistic Graphical Models in Intelligent Systems

Oggetto:

Academic year 2019/2020

Teacher
Prof. Luigi Portinale (Titolare del corso)
Degree course
PhD in Computer Science
Year
1° anno 2° anno 3° anno
Teaching period
Ciclo di incontri
Type
A scelta dello studente
Credits/Recognition
3
Course disciplinary sector (SSD)
INF/01 - informatica
Delivery
Tradizionale
Language
Inglese
Type of examination
Relazione finale
Oggetto:

Sommario del corso

Oggetto:

Course objectives

The course aims at introducing notions and algorithms for Probabilistic Graphical Models (PGM) in Artificial Intelligence (AI). PGMs are the the main AI formalism for dealing with uncertain knowledge and reasoning; they are grounded on both probability calculus and graph theory and represent an effective tool for the construction of intelligent decision support systems. After a short review of probability calculus and of the interpretation of probability, we will discuss different types of PGMs, namely directed models (Bayesian Belief Networks) and undirected models (Random Markov Fields). Both representational and algorithmic issues will be discussed, as well as aspects concerning extensions to dynamic models, sensitivity analysis and learning. Examples of software tools available for the management of PGMs will be presented, by pointing out applications in specific areas.

Oggetto:

Results of learning outcomes

To build a structured probabilistic model
To build a knowledge base supporting uncertain reasoning
To understand independency assumptions in a graphical model
To get a clear meaning of probabilistic inference

Oggetto:

Program

1) Review of probability calculus and of the interpretation of probability

2) Directed PGM: Bayesian Networks
2.1 Modeling and independence
2.2 Inference (exact and approximate)

3) Undirected PGM: Markov Random Fields
3.1 Modeling and independence
3.2 Infwerence (exact and approximate)

4) Dynamic Models: Dynamic Bayesian Networks

5) Decision Models: Influence Diagrams and LIMID

Suggested readings and bibliography

Oggetto:

D. Koller, N. Friedman. Probabilistic Graphical Models: Principles and Techniques, MIT Press

D. Koller. Probabilistic Graphical Models, Coursera Lectures (https://www.coursera.org/specializations/probabilistic-graphical-models)



Oggetto:

Class schedule

DaysTimeClassroom
Monday10:00 - 13:00
Tuesday10:00 - 13:00
Thursday10:00 - 13:00
Thursday14:00 - 16:00

Lessons: dal 10/02/2020 to 13/02/2020

Notes: Sala seminari Dipartimento di Informatica
Seminar Room Computer Science Department

Oggetto:
Last update: 01/02/2020 10:53
Non cliccare qui!