For complex optimization problems like this, why do I never heard about using the significant research done on automatic optimization being used? I would expec

For complex optimization problems like this, why do I never heard about using the significant research done on automatic optimization being used? I would expect each architecture to have particular optimal strategies of memory usage that would be able to be discovered with the machine learning tools used to solve other sorts of optimization problems. I take it from your posts that there isn't a tool you can load up, run it for a few days, then export the most successful strategies it found and then use for parameters obeyed by the system being implemented? Is there a particular reason why? Is the complexity far greater than I'm presuming?