[QIM-8] Add quantum_fisher
#2676
Conversation
Co-authored-by: antalszava <antalszava@gmail.com> Co-authored-by: Edward Jiang <34989448+eddddddy@users.noreply.github.com>
Codecov Report
@@ Coverage Diff @@
## classicalfisher #2676 +/- ##
===================================================
- Coverage 99.59% 99.59% -0.01%
===================================================
Files 253 253
Lines 20375 20384 +9
===================================================
+ Hits 20293 20301 +8
- Misses 82 83 +1
Continue to review full report at Codecov.
|
|
🎉 |
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
Co-authored-by: Romain Moyard <rmoyard@gmail.com>
| @@ -236,3 +237,27 @@ def _make_probs(tape, wires=None, post_processing_fn=None): | |||
| post_processing_fn = lambda x: qml.math.squeeze(qml.math.stack(x)) | |||
|
|
|||
| return [new_tape], post_processing_fn | |||
|
|
|||
|
|
|||
| def quantum_fisher(*args, **kwargs): | |||
There was a problem hiding this comment.
Side note (no action needed): qml.metric_tensor is differentiable, so all the more reason to support differentiability of classical_fisher in the near future.
| - \langle \partial_i \psi(\bm{\theta}) | \psi(\bm{\theta}) \rangle \langle \psi(\bm{\theta}) | \partial_j \psi(\bm{\theta}) \rangle \right] | ||
|
|
||
| with short notation :math:`| \partial_j \psi(\bm{\theta}) \rangle := \frac{\partial}{\partial \theta_j}| \psi(\bm{\theta}) \rangle`. | ||
| It is closely related to the quantum fisher information matrix, see :func:`~.pennylane.qinfo.quantum_fisher` and eq. (27) in `arxiv:2103.15191 <https://arxiv.org/abs/2103.15191>`_. |
There was a problem hiding this comment.
The link to pennylane.qinfo.quantum_fisher doesn't seem to be working. Same with the one in the seealso.
There was a problem hiding this comment.
Yep I also noticed that. Any idea on how to fix it?
| .. seealso:: | ||
| :func:`~.pennylane.adjoint_metric_tensor`; ``quantum_fisher()`` is simply calling :func:`~.pennylane.metric_tensor` with a prefactor of 4. Please refer there for implementation details. | ||
| """ | ||
|
|
There was a problem hiding this comment.
Would be good to add a check here for the shots in tape, such that metric_tensor can be passed if they are finite and adjoint_metric_tensor when they are None. Is that possible for transforms like these?
There was a problem hiding this comment.
That's a really good idea! This isn't currently possible, but we would like to overhaul how shots are architected internally, so that this does become possible.
Currently, shots are stored at the device, not at the transform level
There was a problem hiding this comment.
So for the moment maybe add a boolean variable hardware (or something better) and just do an if-clause?
|
Due to many changes in the base branch started a new PR to simplify things: #2684 |
Context:
Added a dummy function
qml.qinfo.quantum_fisherthat internally callsqml.metric_tensorand multiplies the result by 4 to have the same scaling factor as in eq. (27) in 2103.15191.Additionally, I modified the doc-string of
qml.metric_tensoraccordingly.