Testing phase estimation about grove HOT 8 CLOSED

The representation is as a binary fraction listed from the right. This means that
0001 represents 1/2
0010 represents 1/4
0011 represents 1/4 + 1/2 = 3/4
1000 represents 1/16

Each index to the left is 1/2^i

Does that makes sense @peterwittek ?

from grove.

I think there are two issues here, one is related to #44 and the other is some change to pyquil between late November and now. I'm trying to find the pyquil commit causing the problem.

from grove.

Thanks @kmckiern ! This appears to be fixed.

from grove.

@JansenZhao thanks for raising the issue. That is concerning. I wonder if this is related to issue #44?

from grove.

@ncrubin Thanks. Most likely, but I am not so clear what was meant by "What phase convention did we choose so the QFT is correct?" in #44.

from grove.

I've taken a bit of a look into this and also think that we need to confirm the convention.
When I run the following:

import numpy as np
import scipy.linalg
from grove.alpha.phaseestimation.phase_estimation import phase_estimation
import pyquil.api as api

qvm = api.QVMConnection()

phase = 3/4
Z = np.asarray([[1.0, 0.0], [0.0, -1.0]])
Rz = scipy.linalg.expm(1j*Z*2*np.pi*phase)
p = phase_estimation(Rz, 8)
print(qvm.run(p, range(8), 1))
print(qvm.wavefunction(p))

I get this output:

[[0, 0, 0, 0, 0, 0, 1, 0]]
(1+0j)|001000000>

I would expect

[[0, 0, 0, 0, 0, 0, 1, 1]]
(1+0j)|110000000>

In general if I run with phase of x then I seem to get a bit representation of 1-x and the wavefunction is shifted to the right by one qubit from what i would expect
e.g. for phase=1/16 I get:

[[0, 0, 0, 0, 1, 1, 1, 1]]
(1+0j)|011110000>

from grove.

I'm also getting non-deterministic results for phase=1/8. I would expect that because 1/8 can be exactly represented as a binary fraction then it should give exact and deterministic results? @ncrubin ?

from grove.

Thanks, it works fine! What's the convention for the binary encoding of the phase?

from grove.