Machine Learning (ML) encompasses a large and growing number of methods and concepts. On the one hand, these are grouped along five prediction paradigms namely classification, regression, reinforcement-learning, clustering and association-rule mining. On the other hand, ML methods can be categorized based on the analytics step they address, from data preprocessing and dimensionality reduction to model-selection and evaluation.
There are two viable approaches to teaching ML concepts. The traditional approach is to introduce ML methods and concepts first and then applying them to real world problems. Because of the vast number of problems and approaches encapsulated by ML, a significant amount of time must be spent studying before students can contribute solutions to the academic and industrial worlds. The alternative we pursue is to
- First introduce a real world problem along with the associated data,
- Introduce the elementary theoretical and practical concepts required for solving the given problem, and, finally,
- Encourage students to further investigate the realm of ML by creating prediction competitions based on the problem.
The following datasets are currently tackled by the MachineLearnAthon:
Want Us to Solve Your Problem?
We welcome companies to cooperate with us to solve pressing machine learning issues. We are happy to incorporate new datasets and associated use-cases in our teaching formats. Feel free to contact us!