Leveraging GitHub Classroom and Google Colab for Short-Term Student Machine Learning Team Projects

Abstract

Team projects are an essential part of many university Computer Science curriculums, particularly in the areas of Machine Learn ing and Deep Learning. Large team projects, like Capstone Projects, may be all-encompassing and very intensive, requiring at least a full semester of dedicated work, from project setup to final presentation. However, some courses may not introduce team projects until the second half of the semester, making it more difficult for student teams to establish a unified programming environment and collaborative workflow across personal devices in a shortened time span. Therefore, this paper proposes utilizing GitHub Classroom to create student team repositories for project status updates and submission, in tandem with Google Colab for GPU-dependent machine learning model training. These tools can be directly integrated with each other to provide students with a streamlined and organized workflow that requires minimal project setup and a centralized repository for all project and presentation files. Instructors also benefit from the combination of GitHub Classroom and Google Colab which enables them to spend more time on course content and less time on specific project problems. As an additional benefit, the results of this research indicate that student participation and attendance in class may improve by incorporating a student team project using these tools.