Encoding complex-valued data

[kahaaga]

See #JuliaDynamics/Associations.jl#365.

It would be nice to have a few ways of symbolizing complex-valued data, which I don't think we can do at the moment.

[Datseris]

we have ways of symbolizing 2D data though, and complex valued data can be made 2D one column for reals and one for imaginarys. In dynamical systems theory this is also best practice: don't use complex numbers, use 2D data instead.

My vote would be to support this best practice in the docs instead of supporting complex data. In the section about input data we can write this.

[kahaaga]

That is one way of solving the issue. I guess for mutual information, as in the linked discussion, computing it between two complex valued datasets would be the same as computing it between two 2D StateSpaceSets. One would just have to discretize the magnitudes and angles separately for each dataset.