I am trying to design a model using Q#'s machine learning library that takes in two features (real numbers from 0 to 1) and classifies as 0 or 1.
So how do I decide which Rotations and what seeds to use?
ControlledRotation((0, new Int), PauliY, 0)
I tried using this as the rotation gate but it gave an inaccurate model for many different seed values.
I also tried combining gates:
ControlledRotation((0, new Int), PauliY, 0), ControlledRotation((0, new Int), PauliX, 1)
but this also gave an inaccurate model.
There must surely be a way to analyse what kind of gates to use and not just trial and error. What kind of analysis would be appropriate?