Problems when comparing classical and quantum negative phase

[tbsyyjz01]

Hi,

Thank you for sharing your source code for arXiv:1712.05304. I am PhD student in quantum computation and I am running your code to reproduce your results but I meet some problems.

When I try to print out the classical and quantum negative phase, I got different values, for example:

NEGATIVE PHASE (QUANTUM)
[[9.59270790e-11]
[9.59271278e-11]
[9.59271180e-11]
[9.59271114e-11]]
Negative PHASE(classical)
[[0.25784845]
[0.26259779]
[0.23950028]
[0.23258783]]

The only things I changed in your code is some of QAOA part for new version of grove:
full_QAOA = QAOA(self.qvm, qubits=np.arange(n_system), steps=self.n_qaoa_steps, ref_ham=full_mixer_operator, cost_ham=full_cost_operator, driver_ref=state_prep, store_basis=True, minimizer=fmin_bfgs, minimizer_kwargs={'maxiter':50}, vqe_options={'samples': self.n_quantum_measurements}, rand_seed=1234)

Also, another problem is that the CD-1 algorithm you use is for unit values in {0,1} but your quantum model and DATA should have unit values in {-1,1}, is it right? (correct me if I am wrong).

I tried to modify CD-1 algorithm for unit values in {-1,1}. And the exact negative phase calculated meets the CD-1 result but is off from the quantum results. I also tried to directly calculate the distribution of the quantum part by below code:

model_nus, model_gammas, model_para_prog, _ = self.make_unclamped_QAOA()
model_sampling_prog = model_para_prog(np.hstack((model_nus, model_gammas)))
samplings=self.qvm.run_and_measure(model_sampling_prog,range(self.n_visible),10000)

The distribution I sampled from the quantum is different from what I calculated exactly. I change the steps to 2 and beta to 1(I am not sure here, but beta = 2 still gives wrong distribution) (this is 2 visible units and 2 hidden units RBM without biases)

By beta=1, one example:
quantum: array([0.19426, 0.19714, 0.07278, 0.53582])
exact: array([0.1444419, 0.3555581, 0.3555581, 0.1444419])
By beta=2, one example:
quantum: array([0.10238, 0.10662, 0.01528, 0.77572])
exact: array([0.1444419, 0.3555581, 0.3555581, 0.1444419])

Is there something I missed? Any suggestions on how to get the right quantum distribution would be very appreciated.

Thanks.