Yes this exactly. The Ptolemaic system is an adhoc form of Fourier analysis (circles upon circles). These are universal approximators (another universal class

Yes this exactly. The Ptolemaic system is an adhoc form of Fourier analysis (circles upon circles). These are universal approximators (another universal class are the neural networks) of (preferably smooth) periodic functions. So with enough epicycles one can fit any complicated periodic motion. But it will not help extrapolate the motion of Pluto from the motion of Jupiter, it's a data fit description, not a causal description. This is also why I worry about the recent trend of using DNNs in astronomy. We may go Ptolemaic with them.