3. Learning

3.1 Introduction

Learning, a fundamental characteristic of intelligence, is an integral part of most AI systems, enabling them to optimize their performance.

Machine learning is a very active and multifaceted discipline, depending on the underlying models. All methods involve the existence of large databases of examples and sometimes counter-examples, usually duly labeled, on which learning is based. A system learns from each example, with the idea of being able to generalize its behavior to new cases not yet encountered, thanks to the good properties of the models learned.

There are two main types of learning in AI: symbolic learning and numerical learning. We summarize the main characteristics of these methods below.