How can the transparency and friendliness factor of Machine Learning algorithms be improved?
Any algorithm can be seen as a black box with Input, Settings and Output. This project focusses on Machine Learning Algorithms, which are notoriously difficult to grasp and seem to work by magic. We investigate interactive visual ways to make the black boxes a little bit transparent by allowing users to experiment and “play” with various Input datasets and various parameter Settings. The effects on the Output and on the algorithm performance are immediately observable, hopefully creating new insights on how the machine learning logic operates. In many cases, even the steps of the algorithm can be made visible by animations and simulations.
The target user group is people trying to get familiar with machine learning algorithms for the first time, like for instance students in the Data Science specialisation semesters.