A kötelező előtanulmányi rend az adott szak honlapján és képzési programjában található.

The main objective of the course is to provide a brief introduction to the field of artificial intelligence. The main areas covered by the course include: (1) the problem of expressing intelligent behaviour by computational models, (2) the analysis and application of formal and heuristic methods of artificial intelligence, (3) methods and problems of practical implementations.

1. Introduction: AI problems, intelligence and fundamental issues, engineering approach, history.

2. Analysis of a sample problem. How we manage information. What is needed if the task is non-trivial, but also not impossible. Steps of correct abstraction. What do we gain, what do we give up for it? What are the pitfalls of a given solution?

3. Intelligent systems design: agents, components, environments, architecture and program, search space and basic agent types (behaviour), what to expect inside an agent. What does it mean to be intelligent?

4. Problem solving by search: what are the overall algorithms of intelligent systems, basic mathematical abstractions. How to creatively apply the algorithms we have learned so far to enhance intelligence.

5. The role of paradigm shifts - problem solving by constraint satisfaction. Problem solving in a multiagent environment - search in a hostile environment.

6. The basic component of intelligence - knowledge. Formalising knowledge with logic. What does it mean to reason using logic? There are several forms of logic, how do they differ, what do they provide?

7. Knowledge engineering, logical description of agents, and problem solving by logical inference. Paradigm shift for scaling up.

8. Making plans when everything is going well and when nothing is going well.

9. Intelligence in the real world - incomplete, uncertain and changing knowledge: uncertainty and probability calculation. Probabilistic graphical models, Bayesian networks. Inference in Bayesian networks.

10. Managing temporal knowledge. Rationality and utility. Intelligence as the ability to make rational decisions. Markov decision process.

11. The basic mechanism of intelligence - learning. Basic concepts, basic tasks. Decision tree learning. Learning logical hypotheses.

12. Learning neural networks. Basics of deep neural networks.

13. Learning Bayesian network structures. The main concepts of kernel machines.

14. Reinforcement learning. Q-learning. Deep reinforcement learning.

15. Recommender systems.

16. Problems of multi-agent systems.

Two tests:  a midterm and an endtermtest (MTs) (at a different timeslot than the lecture). The minimum level required for both tests is 40-40%. During the semester, a timed homework assignment is given consisting of several parts, which can be retrieved from an appropriately designed homework server. The schedule for the assignment is available on the assignment homepage. Performance is evaluated based on the sum of the two test scores and the score obtained from the homework (midterm score +endterm score + homework score).

A grade other than unsatisfactory requires a minimum score of 40% on the two tests and 40% of the maximum total score for the semester: max(midterm score)+max(endterm score) +max(homework score).

According to the TVSZ*. Each test can only be corrected once. Late submission of homeworks is possible until the end of the "retake" week.

(* CODE OF STUDIES AND EXAMS OF BME) 

Stuart Russell and Peter Norvig: Artificial Intelligence in Modern Approaches. 

Additional course material is available on the Moodle page of the course.