Hadamard gradient transform (#3625)

* Draft * Trainable params * Working prototype for RX RY RZ * Add op_id * Change tape p idx * Multiple measurements support * Probs * Rot added * Rename * Indep params * Add tests * probs working * Test expval and probs * CR Ising and black * Typos * Tests aux wires * Decomposition * Small changes * Small changes * More black * Pylint tests * Pylint tests * Fix test * Typo * Update * Apply suggestions from code review Co-authored-by: David Wierichs <david.wierichs@xanadu.ai> Co-authored-by: Matthew Silverman <matthews@xanadu.ai> * Apply black * From review * From review * Test PauliRot * pylint tests * Fix multi measurement and multi tape * Doc * Changelog * Change test name * Fix doc * Apply suggestions from code review Co-authored-by: David Wierichs <david.wierichs@xanadu.ai> * Review * Update doc/releases/changelog-dev.md Co-authored-by: Isaac De Vlugt <34751083+isaacdevlugt@users.noreply.github.com> * Pylint * Add to doc * Apply suggestions from code review Co-authored-by: Matthew Silverman <matthews@xanadu.ai> * Note * Add init expand * Update pennylane/gradients/hadamard_gradient.py Co-authored-by: David Wierichs <david.wierichs@xanadu.ai> * Black --------- Co-authored-by: David Wierichs <david.wierichs@xanadu.ai> Co-authored-by: Matthew Silverman <matthews@xanadu.ai> Co-authored-by: Isaac De Vlugt <34751083+isaacdevlugt@users.noreply.github.com>
PennyLaneAI · Feb 3, 2023 · e412162 · e412162
1 parent 30b58bf
commit e412162
Show file tree

Hide file tree

Showing 9 changed files with 1,569 additions and 2 deletions.
diff --git a/doc/releases/changelog-dev.md b/doc/releases/changelog-dev.md
@@ -68,6 +68,28 @@
 
 <h4>Always differentiable 📈</h4>
 
+* The Hadamard test gradient tranform is now available via `qml.gradients.hadamard_grad`.
+  [#3625](https://github.com/PennyLaneAI/pennylane/pull/3625)
+
+  `qml.gradients.hadamard_grad` is a hardware-compatible transform that calculates the 
+  gradient of a quantum circuit using the Hadamard test. Note that the device requires an 
+  auxiliary wire to calculate the gradient.
+
+  ```pycon
+  >>> dev = qml.device("default.qubit", wires=2)
+  >>> @qml.qnode(dev)
+  ... def circuit(params):
+  ...     qml.RX(params[0], wires=0)
+  ...     qml.RY(params[1], wires=0)
+  ...     qml.RX(params[2], wires=0)
+  ...     return qml.expval(qml.PauliZ(0))
+  >>> params = np.array([0.1, 0.2, 0.3], requires_grad=True)
+  >>> qml.gradients.hadamard_grad(circuit)(params)
+  (tensor([-0.3875172], requires_grad=True),
+   tensor([-0.18884787], requires_grad=True),
+   tensor([-0.38355704], requires_grad=True))
+  ```
+
 * The gradient transform `qml.gradients.spsa_grad` is now registered as a
   differentiation method for `QNode`s.
   [#3440](https://github.com/PennyLaneAI/pennylane/pull/3440)

diff --git a/pennylane/gradients/__init__.py b/pennylane/gradients/__init__.py
@@ -47,6 +47,7 @@
     param_shift_cv
     param_shift_hessian
     spsa_grad
+    hadamard_grad
 
 Custom gradients
 ^^^^^^^^^^^^^^^^
@@ -317,6 +318,7 @@ def my_custom_gradient(tape, **kwargs):
 from . import parameter_shift_hessian
 from . import finite_difference
 from . import spsa_gradient
+from . import hadamard_gradient
 
 from .gradient_transform import gradient_transform, SUPPORTED_GRADIENT_KWARGS
 from .hessian_transform import hessian_transform
@@ -327,6 +329,7 @@ def my_custom_gradient(tape, **kwargs):
 from .vjp import compute_vjp, batch_vjp, vjp, compute_vjp_multi_new, compute_vjp_single_new
 from .jvp import batch_jvp, jvp, compute_jvp_multi, compute_jvp_single
 from .spsa_gradient import spsa_grad
+from .hadamard_gradient import hadamard_grad
 
 from .hamiltonian_grad import hamiltonian_grad
 from .general_shift_rules import (

diff --git a/pennylane/gradients/gradient_transform.py b/pennylane/gradients/gradient_transform.py
@@ -40,6 +40,7 @@
     "shots",
     "strategy",
     "validate_params",
+    "device_wires",
 ]