averaging ROC curve for neural networks after some iterations

How to average several ROC curves and their related AUC for Neural Network in Matlab (both vertical averaging and threshold averaging to plot ROC)? In fact, ROC curves belongs to different iterations of a loop. Regards,

[Greg Heath]

I don't understand. What differentiates one ROC curve from another that make the average meaningful?

Answer 1

[Greg Heath]

There is nothing meaningful in averaging ROC curves obtained from different stages in training.

Hope this helps.

Thank you for formally accepting my answer

Greg

[arash ebrahimi]

Thanks for your answer, but according to the following article it seems it is meaningful to do so. Furthermore, the second link is the related code, probably, written in Python. However, I need the Matlab code.

https://ccrma.stanford.edu/workshops/mir2009/references/ROCintro.pdf

https://github.com/hgascon/PALMS/blob/master/roc.py

regards,

[Greg Heath]

I think you misunderstood 2 significant points.

1. The averaging is not the averaging of the ROC curve for a single classifier in different stages of training.

2. JAVA as well as PYTHON code is available for download. Therefore, there may be a better chance for you to obtain a translation

Greg

Answer 2

[AJ Rivera]

Read this paper.

http://www.lcc.uma.es/~jja/recidiva/048.pdf