zbMATH — the first resource for mathematics

Evolutionary algorithms in theory and practice. Evolution strategies, evolutionary programming, genetic algorithms. (English) Zbl 0877.68060
New York, NY: Oxford Univ. Press. xii, 314 p. (1996).
The evolutionary algorithms paradigm originates from an idea to represent the process of solving problem as step by step evolutionary process of some abstract agents evolving to the better solutions. It is based on an organic model of collective randomized learning process within a population of individuals interacting with the environment. Each individual contains some knowledge about the “laws” of the environment and also represents a search point in the space of solutions to a given problem. The starting population evolves towards better regions of search space by means of randomized processes of recombination, mutation and selection.
The reviewed book gives an unified mathematical representation of evolutionary algorithms. The unification, which is the main fundamental contribution by the author, consists in the representation of three compound parts of this computational paradigm, genetic algorithms, evolution strategies and evolutionary programming, as instances of one generalized evolutionary algorithm.
The first part of the book deals with the development of the general functional description of evolutionary algorithms as well as with its biological motivation. The unified definition makes possible the comparison of genetic algorithms, evolution strategies and evolutionary programming according to the criteria of speed and reliability of search. The comparison results are illustrated on five artificial test functions. These functions are sphere model, stop function, Ackley function, functions after Fletcher and Powell, fractal function.
Part 2 of the book concentrates on extending genetic algorithms. Selection mechanisms, mutations and experiments with meta-evolution are mathematically described.
In the appendices some data for the Fletcher-Powell function and from selection experiments is given and also software for genetic algorithms, evolution strategies and evolutionary programming is reviewed.
The book can be useful to researches in the theory of evolutionary algorithms.

68W10 Parallel algorithms in computer science
68-01 Introductory exposition (textbooks, tutorial papers, etc.) pertaining to computer science
68T05 Learning and adaptive systems in artificial intelligence