Evolving Source Code: Object Oriented Genetic Programming in .NET Core

Tools

Speakman, John (2019) Evolving Source Code: Object Oriented Genetic Programming in .NET Core. In: 10th AI and Games Symposium, 16th-18th April 2019, Falmouth University.

[thumbnail of Abstract. Object Oriented Genetic Programming (OOGP) is a method of Genetic Programming (GP) which gives access to standard language libraries, iteration and object-oriented method calls. The implementation of OOGP in this paper shows the automatic gener]
Preview
 Text (Abstract. Object Oriented Genetic Programming (OOGP) is a method of Genetic Programming (GP) which gives access to standard language libraries, iteration and object-oriented method calls. The implementation of OOGP in this paper shows the automatic gener)
Evolving Source Code.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial.
Download (261kB) | Preview

Abstract / Summary

Abstract. Object Oriented Genetic Programming (OOGP) is a method of Genetic Programming (GP) which gives access to standard language libraries, iteration and object-oriented method calls. The implementation of OOGP in this paper shows the automatic generation of retrievable C# files, following standard C# coding conventions with potential access to the entire C# library, derived from a genetic sequence. This new implementation utilises .net Core Roslyn, using reflection, which allows for retrievable, runtime execution and unloading of dynamically generated C# files with scope control in a modern server environment. Experiments were performed on unit tests to validate the algorithms ability to solve simple programming tasks and generate functional, plain text code.
This is a new prototype designed to eventually act as the main Artificial Intelligence controller for a novel, behaviourally adaptive, Artificial-Life simulation. The design taken in the development of this algorithm stems from a requirement for a high potential variation in behaviour, processing efficiency in a server environment per iteration through generated code and low a minimal number of generations.

Item Type: Conference or Workshop Item (Paper)
Subjects: Computing & Data Science
Department: Games Academy
Related URLs:
Depositing User: John Speakman
Date Deposited: 14 Sep 2021 07:12
Last Modified: 08 Aug 2024 09:27
URI: https://repository.falmouth.ac.uk/id/eprint/4343
View Item View Record (staff only)