lesson04 script notebook comments

[elcorto]

These are small and don't need to be addressed immediately. Just want to keep them somewhere.

• [4] I'd use == 0 rather than < 1, so df.species_.cat.codes == 0, only here maybe use >:

  df["not_adelie"] = (df.species_.cat.codes >= 1).astype(np.int8)

  but even then using != 0 is probably more general.

• [16] the output of KNeighborsClassifier.predict_proba() has shape
  (n_queries, n_classes), and n_classes=2 (is_adelie 0 or 1), so I'd use
  index "1" instead of "-1" to stress that there are two classes only and that
  column 1 corresponds to class 1 (=true). Also used here
  https://scikit-learn.org/stable/modules/model_evaluation.html#binary-case

  pos_pred_probs = knn.predict_proba(X_test)[:,1]