Chemistry quantum UCCSD enhanced variational forms

[pbark]

What is the expected enhancement?

Extend of the quantum unitary coupled cluster variational form better suited for reduced circuit depth calculations.

[henrik-dreyer]

Hi Panagiotis,

I'm not sure if this is what you mean, but there are in fact a number of variational forms that are deemed more efficient than UCCSD nowadays. Historically, when people realised that UCCSD had many weaknesses, the next big thing was the Hamiltonian Variational ansatz (or Trotterized Adiabatic State preparation (TASP), see this paper). One of UCCSD's problems is that the number of variational parameters grows as N^4, where N is the number of molecular orbitals, which makes it extremely hard for the classical optimizer to converge. The virtue of the TASP ansatz is that you only need a very modest number of variational parameters.

Inspired by your question, I have toyed around with the idea and implemented the TASP Variational form in Qiskit (see this notebook in this repo ).

These days, variational forms are getting even more efficient (adaptive VQE), by determining their functional form on the fly. Of course that does not easily fit into Qiskit's native VariationalForm class.

Is this what you were asking for? If you are interested I can add my TASP variational form to Aqua.

[pbark]

@henrik-dreyer thanks for this. I will definitely have a look and benchmark your implementation against the different variations of UCCSD. My initial thought was the implementation of advanced UCCSD Ansaetze where the list of excitation is truncated according to certain criteria (e.g. spin constrains). You can find more details of this work in PR #746 which is WIP and in arXiV:1911.10864.

[woodsp-ibm]

@pbark I believe this issue was satisfied by #746, which has been merged/closed, so this can be closed now too right?

[pbark]

@woodsp-ibm I agree. Further expansion and continuation of this work can be discussed in #739 also with @Brogis1