Algorithms and Architectures of Artificial Intelligence

This book gives an overview of methods developed in artificial intelligence for search, learning, problem solving and decision-making. It gives an overview of algorithms and architectures of artificial intelligence that have reached the degree of maturity when a method can be presented as an algorithm, or when a well-defined architecture is known, e.g. in neural nets and intelligent agents. It can be used as a handbook for a wide audience of application developers who are interested in using artificial intelligence methods in their software products. Parts of the text are rather independent, so that one can look into index and go directly to a description of a method presented in the form of an abstract algorithm or an architectural solution. The book can be used also as a textbook for a course in applied artificial intelligence. It has been used as a course material for this kind of courses on undergraduate and graduate levels at Tallinn University of Technology and Royal Institute of Technology in Stockholm. Exercises on the subject are added at the end of each chapter. Neither programming skills nor specific knowledge in computer science are expected from the reader. However, some parts of the text will be fully understood by those who know the terminology of computing well.

The algorithms developed in the field of artificial intelligence are often directly applicable in solving various difficult problems and, therefore, they constitute a valuable part of the artificial intelligence that can be used independently of the philosophical and deep theoretical considerations related to the research in this area. Most of the algorithms are presented in this book in a simple algorithmic language. Sometimes, e.g. when describing neurons in artificial neural nets, it has been more expressive to use common mathematical notations for presenting the functionality of entities. Taxonomy of algorithms and methods concerned is discussed at the end of each chapter, summarizing the material presented in it. This taxonomy can be a valuable aid in developing new algorithms by elaborating the presented ones.