Development of Dl Systems - A Data Science Project Approach

Advances in the use of deep learning (DL) have triggered its large-scale use in a wide range of applications. However, using it in critical applications can have unexpected consequences, leading to concerns about its reliability, repeatability, dependability, and maintainability, which leads to more systematic forms of development and quality assurance. The work has approached developing DL systems, seen from the perspective of data science (DS) projects. A comparison of some methodologies used in DS projects was presented. CRISP-DM was highlighted as the most popular in the industry.