Genetic Algorithms in a Visual Declarative Programming

Abstract

mperative languages like Java, C++, and Python are mostly used for the implementation of Genetic algorithms (GA). Other programming paradigms are far from being an object of study. The paper explores the advantages of a new non-mainstream programming paradigm, with declarative and nondeterministic features, in the implementation of GA. Control Network Programming (CNP) is a visual declarative style of programming in which the program is a set of recursive graphs, that are graphically visualized and developed. The paper demonstrates how the GA can be implemented in an automatic, i.e. non-procedural (declarative) way, using the built-in CNP inference mechanism and tools for its control. The CNP programs are easy to develop and comprehend, thus, CNP can be considered a convenient programming paradigm for efficient teaching and learning of nondeterministic, heuristic, and stochastic algorithms, and in particular GA. The outcomes of using CNP in delivering a course on Advanced Algorithm Design are shown and analyzed, and they strongly support the positive results in teaching when CNP is applied.