Resources /

Software engineering challenges of deep learning

    Our research team share key software engineering challenges of deep learning - from development, to production and organisational challenges.

    There are many challenges with building production-ready systems with deep learning components, especially if the company does not have a large research group and a highly developed supporting infrastructure.

    The research paper “Software Engineering Challenges of Deep Learning,” written by Peltarion members Anders Arpteg, Björn Brinne, Luka Crnkovic-Friis and Jan Bosch, identifies and outlines the main software engineering challenges associated with building systems with deep learning components. Seven projects are described to exemplify the potential for making use of machine learning and specifically deep learning technology. Additionally, to clarify the problematic areas in more detail, a set of 12 challenges are identified and described in the areas of development, production and organizational challenges.

    Why read this

    The main focus of the paper is not to provide solutions, but rather to outline problem areas and, in that way, help guide future research. In addition, the outlined challenges also provide guidance on potential problem areas for companies interested in building high-quality deep learning systems.

    The paper was presented at the 44th Euromicro Conference on Software Engineering and Advanced Applications in September 2018.

    Read, print and download the paper here