Question about spectral pitch display

OK, I am aware from reading through the posts that the spectral pitch display is considered "useless" for editing. I'm not here to debate that. My research area involves deep learning models for sounds - and while working on something else I noticed for a particular data set that when I displayed samples from different categories using the spectral pitch display there were striking differences. When I switched to the spectral frequency display the difference between categories remained but were far less pronounced. 

The short version is that I can use these difference to build more accurate deep learning models, IF I knew the specific algorithms used to produce the displays. I can't find that in the documentation which honestly surprised me. If anyone from Adobe happens to review these posts and can provide that information I would be very grateful. Thanks!