Yes, that's the level of degeneracy. ANN is pretty universal thing, you could train a model for nearly anything that have a static correspondence between inputs and outputs. Say, if you calculating sin(x) with some limited precision, say 4 decimal digits, you could train ANN to do that. Efficiency will be bad, but who cares today about efficiency?
It resembles me when DSP become cheap and some "progressive" university graduates in electronics began to propose making something like fucking low-pass filter for audio signal (which is a resistor and capacitor) using dedicated DSP IC. :)
Yes, that's the level of degeneracy. ANN is pretty universal thing, you could train a model for nearly anything that have a static correspondence between inputs and outputs. Say, if you calculating sin(x) with some limited precision, say 4 decimal digits, you could train ANN to do that. Efficiency will be bad, but who cares today about efficiency?
It resembles me when DSP become cheap and some "progressive" university graduates in electronics began to propose making something like fucking low-pass filter for audio signal (which is a resistor and capacitor) using dedicated DSP IC. :)