How Data Challenges change the way to learn

on the January 15, 2020

The Grenoble Alpes Data Institute is involved in different training programs. It seeks to promote and support teaching of data related skills using innovative education methods such as data challenges.
The MSIAM (Master of Science in Industrial and Applied Mathematics (MSIAM), MoSIG (Master of Science in Informatics at Grenoble) and SIGMA masters (Signal ImaGe processing Methods and Applications) have tried out a new way to evaluate students for 3 years with data challenges.

During several months groups of students are working together on a real problematic to solve. The first challenge was on audio-visual analysis with INRIA Grenoble, the second one on the air pollution with Atmo Auvergne-Rhône-Alpes and the third on, in progress, on the snow composition with GIPSA-Lab.

Students attend to a first seminar about the context (during a Data Science Seminar) and a second one more technical. But teachers are not helping them on the method to use. They have to do a state of the art, see how similar problems were solved. They are autonomous and free to experiment the methods they want to. And they really like it.

This is a genuine benefit for them because they learn how to work in mixed groups. The problem can be solved with computer science, applied mathematics and signal processing skills, the students being expert in the 3 fields. Each student brings her/his own skills to solve the common problem. For this, they also need to communicate and find a common language.

The SSD master (Master in Statistical and Data Science) has tried out data challenges, but this time to evaluate students for 3 “UE”: biostatistics, reliability and machine learning. In these UE, the standard 2-hours-exams and homeworks were replaced with a data challenge.

Students are working by groups of 3 to 5. They have a whole day to find the best code and submit it on the Codalab plateform. They are evaluated on the method they used more than on the learning score they obtain.

Students are really focused during this intensive day work. They are able to propose better results after this 10 hours than during a semester. Adeline Leclercq-Samson, director/supervisor of this masters has notice a high scientific level. Students are critical and use a real scientific procedure.

In these 2 “learning by doing” examples, data challenges enable students to develop their knowledge but moreover their skills to work in industries or in a research center: communication, group work, stress management…

Published on January 15, 2020