feedback SML chapter

[damian0604]

Lastly, as you mentioned in the first class that you are writing a new textbook and any suggestions will be valuable, I do have some suggestions on the SML part.

1. I think it'll be better (at least for me) to understand the concepts if you can add mathematical models and definitions to it.

2. I feel like some terms will need more definitions to better-understand. For example, the ROC-AUC curve. The definition of FPR and TPR are lacking, but I think some math models can certainly help the students to understand the curve a little better.
   TPR = P(Y_predicted = 1 | Y = 1); FPR = 1- P(Y_predicted = 0 | Y = 0)

3. I think it's also good to include why AUC curve is sometimes a better measurement for a predicted model rather than accuracy of F1 score. From my research and understanding, the AUC curve is better because its assumptions are all Y, not Y_pred. Therefore, it's more accurate than accuracy and F1-score (because precision is partly based on Y_pred).

I hope these suggestions help and thank you for your advice in advance!

Best,

Chih-Hsuan (Ian) Lee