When: March 24, 2021, 11:00 - 12:00
Topic 1: What if(True)? Testing your ML models with metamorphic transformations on Code (Leonhard Applis)
What does a deep neural network, such as CodeBERT, really learn? The metrics are State of the Art - but we have no clue why. With our last project Annibale and me looked into ways to alter the test-set to gain more information on what is happening in that beautiful model of yours. We cover everything your heart desires: Machine Learning, Code, Testing and Statistics. So come and join us!
Topic 2: First Good Issues (Andy Zaidman; work with David Alderliesten)
In order to help newcomers to open source projects identify tasks that are suitable to them and their level of expertise within the project, issues can get the good first issue label on the GitHub platform. In this paper we report on a preliminary investigation of good first issues in terms of how they effective they are for developer onboarding and task completion. We find that, although good first issues are effective at developer onboarding, and developers perceive good first issues as being useful, changes can be made to the types of tasks suggested as good first issues to match the types of initial contributions made by newcomers.