That’s odd.

There is work such as complex valued artificial neural networks…

We must do whatever it takes to derive richer and richer representations in weight space. I fear that lack of empirical observation of Supersymmetry at the LHC is inhibiting work based on the reality that formal methods from Supersymmetry may provide deeper abstractions/representation power in ANNs. Whether or not Supersymmetry is found at the LHC, SU(m|n) group representation or formal methods from Supersymmetry are mathematical structures that may give rise to richer nets.

Real valued networks are ancient..