Tensor Network feature request

[akshat998]

I would like to integrate tensor networks capable of:

• Producing samples that can be used by the existing RBM for training. The tensor network shall be implemented for :

  • POVM Types : Pauli-4, Pauli, Terta, Trine, Psi-2
  • MPS Types : GHZ (along with the option of reading MPS from user-provided files)
  • N qubits
  • With the option of adding local noise

• A tensor networks capable of evaluating quantum fidelity for large qubit systems based on the number of samples

The proposal is based on the paper: Reconstructing quantum states with generative models And is being implemented under the guidance of Dr. Juan Carrasquilla

[Roger-luo]

I don't think this will go into QuCumber. We have a (work in progress) package here:

https://github.com/MelkoCollective/Pepinillo

Please feel free to file us an issue there.

[rgmelko]

Currently we keep the sampling codes separate from the reconstruction codes. For example, Dan's iTensor-based sampling code for his QuCumber study is here:

https://github.com/MelkoCollective/SourKRAUT

We should discuss whether the POVM-based sampling goes under SourKRAUT or Pepinillo. My preference would be SourKRAUT I think.

[akshat998]

Cool. Although, I think we are favouring the idea of combining the tensor networks + models in a new repository.