DIGNEA: A tool to generate diverse and discriminatory instance suites for optimisation domains

Alejandro Marrero, Eduardo Segredo, Coromoto León, Emma Hart

Keywords: Instance generation, Novelty search, Evolutionary algorithm, Optimisation, Knapsack problem

Abstract:

This paper presents DIGNEA, a novel software tool designed to generate diverse and discriminatory instance suites for various optimisation domains. DIGNEA utilizes an evolutionary algorithm-based framework that incorporates novelty search techniques to ensure the generated instances are not only varied but also useful for assessing different solvers. The tool, written in C++, is available as a repository and a Docker image, making it easily adaptable to different domains and solver types. Recently, a Python version has been released to facilitate the adoption by the research community. An application to the Knapsack Problem is showcased, demonstrating its effectiveness in generating meaningful test instances.

Submission Number: 9