This is week 2 of the weekly AI challenge
This is just a PoC. I train a layer to basically evaluate a cubic model over the input vector (it's in one dim here coz i cant be bothered)
And then add the coeffiecients of degrees higher than 1 to the loss in order to combat overfitting. But as you can guess, this may cause us to lose more optimal fits and underfit instead. So I will just fiddle around till it works.