Computer Implementation of Genetic Programming

This chapter presents the computer implementation of the tree-based genetic programming in C# programming language. Since C# is a common object-oriented programming language, with little modification the source code presented in the chapter can be easily transformed into Java or C++ programming languages. The chapter covers all aspects of the implementation: node, chromosome, population, function set, and terminal set class implementations. The chapter is carefully structured, so at the end of the chapter fully working GP computer program will be implemented which can solve regression and multiclass classification problems. The reader should not worry about specific operating system, or development environment, since all code implementations are based on cross-OS and open source integrated development environment visual studio code which can run on Windows, Mac, or Linux.

GPdotNET Open Source Software for Running Genetic Programming

In this chapter, GPdotNET v5 genetic programming tool is presented from the user's perspective. GPdotNET is a computer program for running tree-based genetic programming, and its application is modelling supervised machine-learning-based problems. The chapter contains detailed information on how to use GPdotNET in order to prepare data, setup GP parameters, and to run the GP search algorithm. Since GPdotNET supports all three kinds of supervised machine learning problems, the chapter contains three use cases which demonstrate how to successfully build high quality regression, binary, and classification models. GPdotNET contains export module, where the user is able to export GP model to Excel, R language, and Wolfram Mathematica.

Genetic Programming Applications in Solving Engineering Problems

This chapter explains how to use genetic programming to solve various kinds of problems in different engineering fields. Here, three applications, each of which relevant to a distinct engineering field, are explained. First, the chapter starts with the GP application in mechanical engineering. The application analyzes the use of GP in modelling impact toughness of welded joint components. The experimental results of impact toughness represent the input data to build GP models of each welded joint component individually. The second part of the chapter shows how two recent versions of GPdotNET can be satisfactorily used for a binary classification-prediction problem in civil engineering. This application puts forward a new classification-forecasting model, namely binary GP for teleconnection studies between oceanic and heavily local hydrologic variables. Finally, the third application demonstrates how GP could be applied to solve a time series forecasting problem in the field of electrical engineering.

Different Approaches in Genetic Programming

The GP method explained in previous chapters was about the evolution of computer programs represented by monolithic gene (syntax tree). This is the original and most widespread type of GP that is also referred to as tree-based GP. In recent years, new variants of GP have emerged that follow the basic idea of traditional GP to automatically evolve computer programs, but the programs are evolved/represented in different ways. New variants of GP include but are not limited to stack-based genetic programming, linear genetic programming (LGP), Cartesian genetic programming, grammatical evolution (GE), graph-based GP (GGP), context-free grammar (CFGGP), multigene genetic programming (MGGP), and gene expression programming (GEP). Among these variants, main features, evolution of computer programs, and a brief review of engineering applications of MGGP, GEP, and LGP are introduced in this chapter.

Genetic Programming as Supervised Machine Learning Algorithm

This chapter presents the theory and procedures behind supervised machine learning and how genetic programming can be applied to be an effective machine learning algorithm. Due to simple and powerful concept of computer programs, genetic programming can solve many supervised machine learning problems, especially regression and classifications. The chapter starts with theory of supervised machine learning by describing the three main groups of modelling: regression, binary, and multiclass classification. Through those kinds of modelling, the most important performance parameters and skill scores are introduced. The chapter also describes procedures of the model evaluation and construction of confusion matrix for binary and multiclass classification. The second part describes in detail how to use genetic programming in order to build high performance GP models for regression and classifications. It also describes the procedure of generating computer programs for binary and multiclass calcification problems by introducing the concept of predefined root node.

Fundamentals of Genetic Programming

In the living world, all species share one very important property, they receive it right after the birth, and it is called the survival instinct. Since the middle of the twentieth century, scientists have been applying the phenomenon in engineering in order to define computer algorithms which follow the principles of biological evolution of species. Eighty years later, scientists and engineers are still applying the phenomenon in order to solve today's most complex and wide variety of problems. This chapter introduces evolutionary algorithms and motivates the reader to start a journey into genetic programming (GP). The chapter starts with the introduction and detailed insights into GP by describing its main parts and terminology in order to define and mimic biological terms with terms in genetic programming. Then the reader is introduced with the historical evolution of GP and the main and the most popular genetic programming variants, it may find dozens of cited references in it. The chapter continues with detailed introduction on the chromosomes, population, initial and selection methods, main genetic operators, various types of fitness functions, termination criteria, etc. Since GP is processor intensive algorithm, it requires parallel execution to increase its performance which is described at the end of the chapter.