Behavioral Program Synthesis With Genetic Programming

The scope of this book is heuristic program synthesis, where the goal is to automatically generate a program that meets a given set of requirements, provided either as a set oftests or constraints concerning program input and output. The focus of the book is on the generate-and-test approach to program synthesis, of which the primary representative is nowadays genetic programming (GP), a bio-inspired methodology of program induction based on the metaheuristic of evolutionary computation (EC). This book provides a consolidated birds-eye perspective on the approaches to program synthesis that reach beyond the conventional GP template. To accomplish this mission and quickly engage the reader in discussion on the key issues, the book spends relatively little time on introductions, keeping them at minimum. The book covers selected recent development in program synthesis, particularly as seen from the perspective of genetic programming. To ease the navigation, the key concepts are marked with margin notes and listed in the index of terms.

The evolutionary methodology of genetic programming, the program synthesis methodology used in this book, demonstrates that programming can do away without humans. Programs canemerge as an outcome of interactions of a synthesis method with some form of ‘environment’. The final result of such a process may be not different from a program designed by a human. The time has come to seriously consider the possibility of computer systems that autonomously program themselves. The growing body of literature on and achievements of contemporary program synthesis cited in this volume forms a strong evidence for this claim.