Learning objectives
Provide an introduction to modern Artificial Intelligence (AI) with a particular regard for the issues related to the various forms of logic reasoning.
Taking Dublin Indicators into account:
Knowledge and understanding
The course introduces the first concepts related to AI.
Particular emphasis is given to the understanding of the classical
methodologies. The reference text is in Italian, but standard English
terminology is commonly used during the lessons as goodwill to the
consultation of the international scientific literature.
Applying knowledge and understanding
The knowledge presented is always applied to the resolution of specific
problems. The exercises that accompany the course are focused on
solving exercises and problems. Often the solution methods are presented in the
form of an algorithm, developing in students the ability to structure
procedures that are useful in many parts of computer science, and not
only in the study of AI.
Making judgments
The exercises, which are proposed in relation to the theoretical part
presented in class, can be solved individually or in groups. The
comparison with classmates, work at home or in classroom, favors the
development of specific skills in students to enable the explanation of
arguments to fellows and teachers. Often the exercises can be solved in
many different ways and listening to the solutions proposed by other
allows students to develop the ability to identify common structures,
beyond the apparent superficial differences.
Communication skills
The numerous discussions on the different methods to solve problems
allow students to improve communication skills. Specific communication
of AI is also usually used during classes and exercises.
Learning skills
The study of the origins of technological solutions and their introduction
motivated by qualitative and quantitative considerations contributes to
the students’ ability to learn in a deep way and not just superficial and
repetitive. The knowledge acquired is never rigid and definitive, but it is
adaptable to any evolution and change of perspective and context.
Prerequisites
Basic programming skills.
Course unit content
Artificial intelligence and agents.
Problem solving via search.
Games and adversarial problems.
Constraint satisfaction problems.
Logic-based agents.
Planning.
Structured knowledge representation.
Learning.
Neural networks.
Multi-agent systems.
Full programme
Artificial intelligence and agents
Chapters 1 and 2 of the textbook. An introduction to Artificial Intelligence and to the rational agent metaphor.
Problem solving via search
Chapters 3 and 4 of the textbook. Problem solving via search in the state space. Breadth and depth search. Informed search: the A* algorithm. Local search: genetic and evolutionary algorithms.
Games and adversarial problems
Chapter 5 of the textbook. Games via search in the state space: the minimax algorithm and alpha-beta pruning.
Constraint satisfaction problems
Chapter 6 of the textbook. Constraint satisfaction problems (CSPs). CSP solving via backtracking. Types of consistency and arc-consistency. Forward checking and algorithms for local consistency maintenance.
Logic-based agents
Chapters 7, 8 and 9 of the textbook. Propositional logic: clauses and resolution. First order logic and basics of resolution and logic programming.
Planning
Chapter 11 of the textbook. General characteristics of a planning system. The blocks world. STRIPS. Real-world planning: conditional planning and execution control.
Structured knowledge representation
Description logics and structured inheritance networks. Ontologies and their applications to the Semantic Web.
Learning
Chapter 18 of the textbook. Inductive learning: decision trees. Reinforcement learning.
Neural networks
Perceptrons and feed-forward networks. Reinforcement learning and the back propagation algorithm
Multi-agent systems
Cooperative and competitive agents and multi-agent systems. Communication between agents and communicative acts. FIPA and the BDI model (brief introduction to modal logics).
Bibliography
Stuart Russell e Peter Norvig. Intelligenza artificiale: un approccio moderno. UTET Libreria, 1998.
Teaching methods
Classes and laboratories are located at Dipartimento di Matematica e Informatica.
Laboratory exercizes share time slots with classes.
Dates of exams will be available at http://informatica.unipr.it
Meetings with the teacher can be requested via e-mail.
Assessment methods and criteria
Being able to understand and make appropriate use of techniques of modern Artificial Intelligence.
The exam consists of an oral report and a project which can be
accessed only after the oral report. It is possible to take the oral report
several times, but each oral report cancels previous projects.
Other information
- - -
