557658 (1) [Avatar] Offline
#1
In listing D.4, false positives are defined with the following code:

false_positives=((y_pred != y_true) & (y_true==1)).sum()

This seems against what is mentioned immediately above listing D.6: "false positives are the negative examples that you mislabeled as positive"

Shouldn't the code actually be as follows?

false_positives=((y_pred != y_true) & (y_true==0)).sum()

Thank you for this valuable book!
hobs (58) [Avatar] Offline
#2
You are right! Stop the presses! Thanks to your contribution we've corrected it at the last minute and the book should be correct and more clear. We also changed the y_true use in the true_positive and true_negative calculation to be consistent and clear. The nlpia package has an example for appendix D with your name on it (forum id). Thank you so much. It's people like you that will use NLP and ML to make the world a better place.