Teaching Machine Learning to a Diverse Audience: the Foundation-based Approach

How should we teach machine learning to a diverse audience? Is a ‘foundation-based’ approach appropriate or should we switch to a ‘techniques-oriented’ approach? We argue that the foundation-based approach is still the way to go. The foundation-based approach emphasizes the fundamentals first, before moving to algorithms or paradigms. Here, we focus on three key concepts in machine learning that cover the basic understanding, the algorithmic modeling, and the practical tuning. We then demonstrate how more sophisticated techniques like the popular support vector machine can be understood within this framework.