Modify loss function used in classification layer

Asked by Jennifer Voon on 26 Sep 2019

Latest activity Commented on by Kenta Itakura on 28 Sep 2019

4 views (last 30 days)

The default loss function used in classification layer is the crossentrophy. How can I modify or change the loss function used in the classification layer?

2 Comments

Greg Heath on 28 Sep 2019

You will have to define EXACTLYwhat output you want for each class's input.

Greg

Jennifer Voon on 28 Sep 2019

Thank Greg for replying.

I am aware of that, but I am not exactly sure how to ‘define’ the output. Is there any example code that shows how I can do that?

Thank you!

1 Answer

Link

Answer by Kenta Itakura on 28 Sep 2019

Accepted Answer

https://jp.mathworks.com/help/deeplearning/ug/define-custom-classification-output-layer.html?lang=en

Please refer to this document. This document shows an example to change the loss function into sum of squares error.

In my experience, changing the function into other is not easy, though it is technically possible.

2 Comments

Jennifer Voon on 28 Sep 2019

Thank you Kenta!

Kenta Itakura on 28 Sep 2019

my pleasure

You are now following this question

Jennifer Voon (view profile)