Hi everybody,
I have one doubt regarding the Error on the Test and Validation Chart, which according to Jeremy should be a line when using different models (in the procedure to find a proper validation split).
I understand the reason that it should be very close to the line, but my question is: should I expect this line to be 45 degrees (both errors to be the same) as long as my loss metric is an average metric (divided by N - number of samples)?
Thanks,
Mauricio