This is the corresponding optimizer to my blog post
I take in the current and previous step's 0th and 1st order information to fit a cubic polynomial that we can analytically solve for minimum in the cases of oscillations in parameter space.
needed reqs: tf 2.0.0 and numpy