Hello,
That's a strange question. 🤔🤔
What do you mean with 'machine level code'? The source code of SAS itself (often C programming language) is not publicly available.
Which machine learning front-end of SAS are you using and what is your SAS version?
The front-ends (like Enterprise Miner or Visual Data Mining and Machine Learning in Model Studio) are all SAS-code generators and this SAS code can be looked at of course. The training code can be seen, the scoring code, the LOG info, about everything ...
Also the documentation (https://documentation.sas.com/)
gives quite exhaustive info on the algorithms and how they were implemented and what they exactly do (including formulas and equations).
But what you mention (log transformation, ADAM, logistic regression, tanh function) are well-known things. SAS is not re-inventing them but just implementing them from widely accepted theory.
Although I agree that things like nonlinear optimization can be implemented in a good and bad way.
I am inclined to think that SAS implements things / theory in a good way.
Let us know what you are exactly after.
Koen
