Demo has inefficient kernel

The kernel circuit is a little inefficient in this demo.
At the crosspoint of regular ansatz and adjoint ansatz the parametarized layers cancel each other out effectively losing a layer. (So 1 layer actually has no layers and is straight up only feature embedding)

I suppose it can improved by simply embedding the features after the parametarized layer not before e.g. (layer, feature, layer, feature) not (feature, layer, feature, layer)

Sorry, I didn’t know where to post this, or maybe I am wrong, please correct me if I am :slight_smile:

Hey @ilmars_kuhtarskis!

Interesting! I’m lacking a bit of information here, but if you think you can get the demo to run faster or has some redundancies then you can make a pull request or an issue on the following Github repository: