I was always curious as to why we don't encode ML parameters and activations using complex numbers. A dot product between two complex numbers naturally encodes

I was always curious as to why we don't encode ML parameters and activations using complex numbers. A dot product between two complex numbers naturally encodes confidence in the result in the magnitude. I blame numpy