Adds Single Excitation operations #1121

ixfoduap · 2021-03-03T20:57:20Z

Context:
Current tools for constructing quantum circuits for quantum chemistry are inefficient. By relying on fermionic-to-qubit mappings to define operations and circuit decompositions to compute gradients, the resulting circuits for creating single and double excitation operations lead to lengthy computations. Through research work, we've identified that Givens rotations perform the same role as fermionic excitation operations, while allowing for simple analytical gradients and faster implementation in hardware and simulators.

Description of the Change:
This PR adds the two-qubit SingleExcitation operation with support across all native devices. Its analytical gradient requires a decomposition into two gates, SingleExcitationPlus and SingleExcitationMinus, which are also added as part of this PR.

A follow-up PR will also add the four-qubit DoubleExcitation operations.

Benefits:
Much faster implementation of quantum circuits for quantum chemistry.

Note:
The SingleExcitationPlus and SingleExcitationMinus have generators G that satisfy G^2=1 (self-inverse). This means they satisfy the standard parameter-shift rule and no custom gradient formula needs to be passed

.github/CHANGELOG.md

…tations

josh146

Looking great @ixfoduap, especially updating all the jax/tf/autograd devices! I'm guessing this is just a pre-review, since there are no tests yet?

.github/CHANGELOG.md

pennylane/ops/qubit.py

Co-authored-by: Josh Izaac <josh146@gmail.com>

…tations

.github/CHANGELOG.md

pennylane/devices/autograd_ops.py

agran2018

Thanks @ixfoduap!. I have enjoyed reviewing this PR while learning how to add new operations to PL.

agran2018 · 2021-03-18T21:37:17Z

pennylane/devices/autograd_ops.py

+    """
+    c = np.cos(phi / 2)
+    s = np.sin(phi / 2)
+    return np.array([[1, 0, 0, 0], [0, c, -s, 0], [0, s, c, 0], [0, 0, 0, 1]])


Docstring declares that the function returns array[complex] but the matrix elements s and c are real, right?

Good catch!

agran2018 · 2021-03-18T21:43:38Z

pennylane/devices/jax_ops.py

+    """
+    c = jnp.cos(phi / 2)
+    s = jnp.sin(phi / 2)
+    return jnp.array([[1, 0, 0, 0], [0, c, -s, 0], [0, s, c, 0], [0, 0, 0, 1]])


Same comment here regarding the type of the returned array, jnp.Tensor[float] ?

agran2018 · 2021-03-18T21:45:22Z

pennylane/devices/tf_ops.py

+    phi = tf.cast(phi, dtype=C_DTYPE)
+    c = tf.cos(phi / 2)
+    s = tf.sin(phi / 2)
+    return tf.convert_to_tensor([[1, 0, 0, 0], [0, c, -s, 0], [0, s, c, 0], [0, 0, 0, 1]])


probably the same here: tf.Tensor[float]

Note: while true for JAX and autograd above, in this case the tensor will be complex, since phi is cast to complex on line 208. So it should remain tf.Tensor[complex].

Note: while true for JAX and autograd above, in this case the tensor will be complex, since phi is cast to complex on line 208. So it should remain tf.Tensor[complex].

Attention to details! 😄

pennylane/ops/qubit.py

agran2018 · 2021-03-18T22:08:15Z

tests/gate_data.py

+        phi (float): rotation angle
+
+    Returns:
+        array: the two-qubit Givens rotation describing the single excitation


"describing the single excitation" or "describing the operation"

tests/ops/test_qubit_ops.py

ixfoduap · 2021-03-19T13:25:11Z

pennylane/devices/autograd_ops.py

+    """
+    c = np.cos(phi / 2)
+    s = np.sin(phi / 2)
+    return np.array([[1, 0, 0, 0], [0, c, -s, 0], [0, s, c, 0], [0, 0, 0, 1]])


Good catch!

pennylane/devices/autograd_ops.py

pennylane/devices/jax_ops.py

pennylane/devices/tf_ops.py

pennylane/ops/qubit.py

tests/gate_data.py

agran2018

Thanks @ixfoduap. Looks great to me!.

agran2018 · 2021-03-19T16:49:28Z

pennylane/devices/tf_ops.py

+    phi = tf.cast(phi, dtype=C_DTYPE)
+    c = tf.cos(phi / 2)
+    s = tf.sin(phi / 2)
+    return tf.convert_to_tensor([[1, 0, 0, 0], [0, c, -s, 0], [0, s, c, 0], [0, 0, 0, 1]])


Note: while true for JAX and autograd above, in this case the tensor will be complex, since phi is cast to complex on line 208. So it should remain tf.Tensor[complex].

Attention to details! 😄

antalszava

Looks good to me! 💯

antalszava · 2021-03-19T17:51:41Z

pennylane/devices/autograd_ops.py

+    r"""
+    Single excitation rotation with positive phase-shift outside the rotation subspace.


As mentioned in the suggestion from the first iteration, this recurs here and in other files too for the newly added docstrings.

Suggested change

r"""

Single excitation rotation with positive phase-shift outside the rotation subspace.

r"""Single excitation rotation with positive phase-shift outside the rotation subspace.

antalszava · 2021-03-19T17:51:50Z

pennylane/devices/autograd_ops.py

+    r"""
+    Single excitation rotation with negative phase-shift outside the rotation subspace.


Suggested change

r"""

Single excitation rotation with negative phase-shift outside the rotation subspace.

r"""Single excitation rotation with negative phase-shift outside the rotation subspace.

antalszava · 2021-03-19T17:52:02Z

pennylane/devices/autograd_ops.py

+
+

This is a suggestion for previous docstrings too here and in other files.

Suggested change

tests/ops/test_qubit_ops.py

antalszava · 2021-03-19T18:18:54Z

tests/ops/test_qubit_ops.py

+
+    @pytest.mark.parametrize(("excitation", "phi"), [(qml.SingleExcitation, -0.1),


Two things:

Could we parametrize for diff_method here too?

How about testing default.qubit for parameter-shift?

.github/CHANGELOG.md

tests/ops/test_qubit_ops.py

antalszava · 2021-03-19T18:29:19Z

tests/ops/test_qubit_ops.py

+            pytest.skip("JAX support for the parameter-shift method is still TBD")
+
+        jax = pytest.importorskip("jax")
+        from jax import numpy as jnp


How come this import is here and for tf we don't have one? 🤔 Was there an occurrence for TF already before in the file that imported it?

Co-authored-by: antalszava <antalszava@gmail.com>

pennylane/devices/tf_ops.py

…tations

ixfoduap added 6 commits February 24, 2021 09:01

First version of new excitation operations

c211c02

Adds docstrings + interface operations

c9aa6d4

Merge branch 'master' into excitations

49d231e

Adds support across devices and changelog snippet

143cbe3

Adds name to contributors

5f12bbc

Black

6382dc1

ixfoduap added core Core functionality (Operations, QNodes, CircuitGraphs) WIP 🚧 Work-in-progress qchem ⚛️ Related to the QChem package labels Mar 3, 2021

ixfoduap requested review from josh146, agran2018 and trbromley March 3, 2021 20:57

ixfoduap self-assigned this Mar 3, 2021

ixfoduap commented Mar 3, 2021

View reviewed changes

.github/CHANGELOG.md Outdated Show resolved Hide resolved

ixfoduap and others added 4 commits March 3, 2021 15:57

Update .github/CHANGELOG.md

9769875

small changes

0521cd2

Merge branch 'excitations' of github.com:XanaduAI/pennylane into exci…

fa16191

…tations

Merge branch 'master' into excitations

a7c4f97

josh146 reviewed Mar 4, 2021

View reviewed changes

.github/CHANGELOG.md Outdated Show resolved Hide resolved

pennylane/ops/qubit.py Outdated Show resolved Hide resolved

pennylane/ops/qubit.py Outdated Show resolved Hide resolved

pennylane/ops/qubit.py Outdated Show resolved Hide resolved

pennylane/ops/qubit.py Show resolved Hide resolved

ixfoduap and others added 3 commits March 4, 2021 09:41

Apply suggestions from code review

ae018d5

Co-authored-by: Josh Izaac <josh146@gmail.com>

Adds tests and improves docstrings

23e85ed

Merge branch 'excitations' of github.com:XanaduAI/pennylane into exci…

4cd76e7

…tations

ixfoduap commented Mar 4, 2021

View reviewed changes

.github/CHANGELOG.md Outdated Show resolved Hide resolved

Update .github/CHANGELOG.md

882bb77

ixfoduap commented Mar 4, 2021

View reviewed changes

.github/CHANGELOG.md Outdated Show resolved Hide resolved

Update .github/CHANGELOG.md

40b2320

ixfoduap changed the title ~~Adds Single Excitation operations [WIP]~~ Adds Single Excitation operations Mar 4, 2021

ixfoduap removed the WIP 🚧 Work-in-progress label Mar 4, 2021

ixfoduap commented Mar 4, 2021

View reviewed changes

pennylane/devices/autograd_ops.py Outdated Show resolved Hide resolved

pennylane/devices/autograd_ops.py Outdated Show resolved Hide resolved

pennylane/devices/autograd_ops.py Outdated Show resolved Hide resolved

Apply suggestions from code review

3e9f7ba

ixfoduap and others added 2 commits March 18, 2021 17:26

Update .github/CHANGELOG.md

95d9abc

fixes changelog

0c059ea

agran2018 reviewed Mar 18, 2021

View reviewed changes

this should fix it

d85a148

ixfoduap commented Mar 19, 2021

View reviewed changes

ixfoduap and others added 2 commits March 19, 2021 09:31

Apply suggestions from code review

c297f75

fix indentation

434a9ba

ixfoduap requested review from josh146, agran2018 and antalszava March 19, 2021 14:01

agran2018 approved these changes Mar 19, 2021

View reviewed changes

antalszava approved these changes Mar 19, 2021

View reviewed changes

antalszava reviewed Mar 19, 2021

View reviewed changes

tests/ops/test_qubit_ops.py Outdated Show resolved Hide resolved

antalszava reviewed Mar 19, 2021

View reviewed changes

.github/CHANGELOG.md Outdated Show resolved Hide resolved

improvements to docstrings

2476d1f

antalszava reviewed Mar 19, 2021

View reviewed changes

tests/ops/test_qubit_ops.py Outdated Show resolved Hide resolved

antalszava reviewed Mar 19, 2021

View reviewed changes

josh146 and others added 4 commits March 22, 2021 16:29

Merge branch 'master' into excitations

ac05f79

fix docs

9ec2fd3

Merge branch 'master' into excitations

9348192

Apply suggestions from code review

042f2b9

Co-authored-by: antalszava <antalszava@gmail.com>

ixfoduap commented Mar 22, 2021

View reviewed changes

pennylane/devices/tf_ops.py Outdated Show resolved Hide resolved

ixfoduap and others added 4 commits March 22, 2021 14:20

Update pennylane/devices/tf_ops.py

a1ef6f1

changes from code review

1d5fc6b

Merge branch 'excitations' of github.com:XanaduAI/pennylane into exci…

083c61e

…tations

fixes tests

ac42e67

ixfoduap merged commit 8803e1f into master Mar 22, 2021

ixfoduap deleted the excitations branch March 22, 2021 21:11

antalszava mentioned this pull request May 5, 2021

Define decomposition for single and double excitation operations #1275

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adds Single Excitation operations #1121

Adds Single Excitation operations #1121

ixfoduap commented Mar 3, 2021 •

edited

Loading

josh146 left a comment

agran2018 left a comment

agran2018 Mar 18, 2021

ixfoduap Mar 19, 2021

agran2018 Mar 18, 2021

agran2018 Mar 18, 2021

josh146 Mar 19, 2021

agran2018 Mar 19, 2021

agran2018 Mar 18, 2021

ixfoduap Mar 19, 2021

agran2018 left a comment

agran2018 Mar 19, 2021

antalszava left a comment

antalszava Mar 19, 2021

antalszava Mar 19, 2021

antalszava Mar 19, 2021

antalszava Mar 19, 2021

antalszava Mar 19, 2021

		r"""
		Single excitation rotation with positive phase-shift outside the rotation subspace.

	r"""
	Single excitation rotation with positive phase-shift outside the rotation subspace.
	r"""Single excitation rotation with positive phase-shift outside the rotation subspace.

		r"""
		Single excitation rotation with negative phase-shift outside the rotation subspace.

	r"""
	Single excitation rotation with negative phase-shift outside the rotation subspace.
	r"""Single excitation rotation with negative phase-shift outside the rotation subspace.


		@pytest.mark.parametrize(("excitation", "phi"), [(qml.SingleExcitation, -0.1),

Adds Single Excitation operations #1121

Adds Single Excitation operations #1121

Conversation

ixfoduap commented Mar 3, 2021 • edited Loading

josh146 left a comment

Choose a reason for hiding this comment

agran2018 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

agran2018 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

antalszava left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ixfoduap commented Mar 3, 2021 •

edited

Loading