Add decompositions of RX, RY and RZ into one of the other two + Cliffords
#8569
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #8569 +/- ##
=======================================
Coverage 99.43% 99.43%
=======================================
Files 588 588
Lines 62727 62757 +30
=======================================
+ Hits 62373 62403 +30
Misses 354 354 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| @pytest.mark.parametrize("op", ALL_OPERATIONS) | ||
| def test_assert_valid(self, op): | ||
| kwargs = SKIP_ASSERT_VALID.get(type(op), {}) | ||
| if kwargs is True: | ||
| pytest.skip() | ||
|
|
||
| qml.ops.functions.assert_valid(op, **kwargs) |
There was a problem hiding this comment.
Thanks for adding this!
|
|
||
| add_decomps(RY, _ry_to_rot, _ry_to_rz_rx) | ||
| def _ry_to_rx_cliff_resources(): | ||
| return {change_op_basis_resource_rep(adjoint_resource_rep(qml.S), qml.RX, qml.S): 1} |
There was a problem hiding this comment.
This one checks out!
There was a problem hiding this comment.
(checked this one by hand...)
|
|
||
|
|
||
| def _rx_to_rz_cliff_resources(): | ||
| return {change_op_basis_resource_rep(qml.Hadamard, qml.RZ, qml.Hadamard): 1} |
There was a problem hiding this comment.
I know Hadamard transforms between X and Z bases.
| @register_resources(_ry_to_rz_cliff_resources) | ||
| def _ry_to_rz_cliff(phi, wires: WiresLike, **__): | ||
| qml.change_op_basis( | ||
| qml.Hadamard(wires) @ qml.adjoint(qml.S(wires)), |
There was a problem hiding this comment.
This one checks out!
There was a problem hiding this comment.
wires = 0
phi = 1.5 * 3.14
matrix = qml.matrix(qml.S(wires) @ qml.Hadamard(wires) @ qml.RZ(phi, wires) @ qml.Hadamard(wires) @ qml.adjoint(qml.S(wires)))
ry = qml.matrix(qml.RY(phi, wires))
assert np.allclose(matrix, ry)
| @register_resources(_rz_to_ry_cliff_resources) | ||
| def _rz_to_ry_cliff(phi, wires: WiresLike, **__): | ||
| qml.change_op_basis( | ||
| qml.S(wires) @ qml.Hadamard(wires), |
There was a problem hiding this comment.
Checks out.
wires = 0
phi = 1.5 * 3.14
matrix = qml.matrix(qml.Hadamard(wires) @ qml.adjoint(qml.S(wires)) @ qml.RY(phi, wires) @ qml.S(wires) @ qml.Hadamard(wires))
rz = qml.matrix(qml.RZ(phi, wires))
assert np.allclose(matrix, rz)
comp-phys-marc
left a comment
comp-phys-marc
left a comment
There was a problem hiding this comment.
All the decompositions look right!
dwierichs
requested review from
JerryChen97
and removed request for
JerryChen97
|
Does this PR affect ftqc somehow? https://github.com/PennyLaneAI/pennylane/actions/runs/19030113144/job/54341829822?pr=8569 |
doc/releases/changelog-dev.md
Outdated
|
|
||
| <h3>Improvements 🛠</h3> | ||
|
|
||
| <<<<<<< HEAD |
There was a problem hiding this comment.
Oops... looks like a bad merge here.
There was a problem hiding this comment.
I believe this is a GH issue. Recently somehow you can't trust the GH entry point from the notifications any more; we have to go back to the PR and click the files section to see the correct diff
There was a problem hiding this comment.
Basically their problem made you see and comment on an outdated part:
doc/releases/changelog-dev.md
Outdated
|
|
||
| <h3>Breaking changes 💔</h3> | ||
|
|
||
| <<<<<<< HEAD |
Context:
The Clifford group by definition can transform Paulis to each other. In particular, we can conjugate
RX,RYandRZby appropriate Cliffords to get either one of the other rotations.Description of the Change:
Add decompositions within the graph-based system that implement this. It amounts to 6 rules (
RX->RYRX->RZ,RY->RXRY->RZ,RZ->RXRZ->RY).Also, add an
assert_validtest to all standard ops.Benefits:
More versatile decomposition, in particular in the context of phase polynomials which prefer
RZoverRXorRYrotations.Possible Drawbacks:
N/A
Related GitHub Issues: