With machine learning I wonder if there could exist something like smart statistics.
For example. Does there exist smart inverse statistics?
That is. Correcting input data by setting up several statistical conditions. Meaning the results should have the correct statistical output based on corrected input.
If you know how to correct than it becomes easier to generate probable input data. Just add and correct ?random input.
Then I wonder if the sound turbulence makes could be treated like audio statistics. I guess the turbulence solution could be set generated (corrected from random) from the condition that it should sound as the recognized type of turbulent noise.