If you expose a neural network to enough empirical molecular orbitals, it’s going to notice ‘patterns’ that are better than using say the radial solution of hydrogen (a gaussian). Picking chinese faces out of a crowd of billions is actually trickier.
“If you expose a neural network to enough empirical molecular orbitals, . . . “
But how do you get the data? Measurement is a problem.
In a way, this is inevitable.