1
$\begingroup$

I've used Classify[...] for a two class image classification problem which has yielded pretty good results. I let Mathematica select an algorithm for me and it picked a GBT model, everything looked good and the algorithm performed well.

I put the whole thing away for a while, but now it's come time to publish and although it's a small part of the study I feel I should know what loss function Mathematica chose for the model. I figured this would be easily accessible but I can't seem to find a citation or mention of this anywhere. I'm fairly certain it must use cross entropy, but "fairly certain" obviously isn't going to get past a reviewer.

If anyone has any idea, or knows how to get this information out of the model after training, I'd be very grateful if you could share your insights.

$\endgroup$
2
  • $\begingroup$ Try evaluating List @@ classifier to see the information stored inside. That should reveal some information about the loss function as well. I'm going to go out on a limb and guess that it's mean crossentropy for a classification task. $\endgroup$ May 17 '21 at 21:11
  • $\begingroup$ This worked perfectly, and your guess was on the money. Thanks very much! If you want to post this as an answer I'll accept it. $\endgroup$
    – dangerfeld
    May 18 '21 at 13:11
0
$\begingroup$

Try evaluating List @@ classifier to see the information stored inside. That should reveal some information about the loss function as well. I'm going to go out on a limb and guess that it's mean cross-entropy for a classification task.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.