Knowledge representation, reasoning and learning are key aspects of the Artificial Intelligence field. This course will focus on the first two components, however, we will also briefly address the main ideas and concepts in machine learning.
Knowledge is a pivotal aspect when it comes to designing knowledge-based systems. Actually, it is difficult to imagine an intelligent system operating without any knowledge, either provided by human experts or derived from historical data. In recent years there has been an increasing interest in developing intelligent systems merging those approaches, therefore leading to hybrid intelligent systems. We have reached this point after understanding that neither pure knowledge-based systems nor automatic systems are the silver bullet.
The content of this course can be divided into three major topics: theoretical fundamentals, symbolic knowledge representation and sub-symbolic knowledge representation. These topics are not independent of each other. The theoretical concepts elaborate on general concepts such as artificial intelligence, knowledge, representation, thinking, and reasoning. Many of these concepts inherently have a philosophical connotation. We will also cover aspects related to solutions we intend to find, their taxonomy, heuristic search algorithms that allow exploring a complex solution space in a reasonable amount of time. Towards the end of the first part of the course, we will discuss a procedure to thinking, which is the fancy we (mathematicians, logicians and computer scientists) have given to the logical entailment procedure.
In the second part of the course, we will further dig into the symbolic knowledge representation and reasoning. The Prolog programming language will be our target. This language could be seen as a computational implementation of the logical entailment procedure to manipulate symbolic knowledge structures. The symbolic artificial intelligence in its pure state!
The last part of this course will be devoted to exploring some relevant sub-symbolic knowledge representation and reasoning models. The selected models are based on Fuzzy Cognitive Maps, which are knowledge-based recurrent neural networks to model and simulate complex systems. Here we will go briefly on the concept of machine learning. Furthermore, we will discuss different approaches to reason in presence of uncertainty.