Restricted Boltzmann machine for Quantum state tomography

[Jufa]

TL;DR Summary

I would like to know how one would carry out quantum tomography from a quantum state by means of the restricted Boltzmann machine. For the sake of simplicity we could choose a 1-qubit system

I'm struggling with my Final Degree Project. I would like to perform a quantum simulation and perform quantum tomography for a single-qubit using a resrticted Boltzmann machine. In order to do so I'm trying to follow the recipe in the paper "Neural Network quantum state tomography, Giacomo Torlai et al.", but I fail to understand it.
How many visible neurons are needed?
How I know which is the value of the visible neurons based on the measurments performed?
These two questions are the main ones.
Could someone please help?
Thanks a lot in advance.

 

[Jufa]

Anyone? I'm quite lost ans frustrated, just a little of debate would extraordinarily help.

 

[PeterDonis]

Do you have a link to the paper you referenced? If so, please provide it. That makes it easier for people to respond, since they don't have to go try to find it on their own.

 

[Jufa]

Sure, here it goes the link:
https://www.labxing.com/files/lab_publications/2278-1524663501-3fRuMVpV.pdf

Thanks for responding.

 

[Jufa]

In addition I don't see how the generalisation of this method to mixed states should be performed.

 

[Jufa]

I think I have finally come up with the two answers to the questions I made. If only someone could verify I'm right it would be very helpful:
How many visible neurons are needed?
-The answer is that we need so many visible neurons as qubits has the system. This way every possible component of the system (in a certain basis) can be reffered as a unique state of the visible neurons.
-How I know which is the value of the visible neurons based on the measurments performed?
Given a certain basis, every result of an experiment performed in that basis becomes an input vector, i.e., a certain state of the visible layer.