Testing Statistical Hypothesis for Phd-students also contains the Theory of Regression. We have a first draft of lecture notes and we will soon add more. For many Machine Learning queetions, it is highly educational to study in detail the mathematical theory of high dimensional l2-regression: since this is one of the few instances where optimal solutions can be calculated explicitly, and hence one can understand exactly what is happening. Also, many Machine Learning practitioners forget optimality theory for statistical hypothesis. When one knows what the optimal solution is, there is no need for tousands of papers, trying all kind of wierd methods. Of course, sometimes the issue is that the model in the real world does not correspond exactly to the theoretical model.