Towards a Standard-based Domain-specific Platform to Solve Machine Learning-based Problems

• Main Authors: Vicente García-Díaz, Jordán Pascual-Espada, Cristina Pelayo G-Bustelo, Juan Manuel Cueva-Lovelle
• Format: Article
• Language: English
• Published: Universidad Internacional de La Rioja (UNIR) 2015-12-01
• Series: International Journal of Interactive Multimedia and Artificial Intelligence
• Subjects: Artificial Intelligence; DSL; Machine Learning; MDE; Xtext
• Online Access: http://www.ijimai.org/journal/node/855

Machine learning is one of the most important subfields of computer science and can be used to solve a variety of interesting artificial intelligence problems. There are different languages, framework and tools to define the data needed to solve machine learning-based problems. However, there is a great number of very diverse alternatives which makes it difficult the intercommunication, portability and re-usability of the definitions, designs or algorithms that any developer may create. In this paper, we take the first step towards a language and a development environment independent of the underlying technologies, allowing developers to design solutions to solve machine learning-based problems in a simple and fast way, automatically generating code for other technologies. That can be considered a transparent bridge among current technologies. We rely on Model-Driven Engineering approach, focusing on the creation of models to abstract the definition of artifacts from the underlying technologies.