Add k-UpCCGSD Template #1743

obliviateandsurrender · 2021-10-12T14:07:33Z

Context:
Added a new template kUpCCGSD, which implements a unitary coupled cluster ansatz with generalized singles and pair doubles excitation operators, proposed by Joonho Lee et al. in arXiv:1810.02327

Description of the Change:
Adds the kUpCCGSD class implementing the template and the relevant test cases.

Benefits:
Extends the currently available ansatz templates for quantum chemistry in PennyLane.

Possible Drawbacks:
No drawbacks.

Related GitHub Issues:
None.

codecov · 2021-10-12T14:21:43Z

Codecov Report

Merging #1743 (4458507) into master (06f2fe8) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #1743   +/-   ##
=======================================
  Coverage   98.91%   98.91%           
=======================================
  Files         206      207    +1     
  Lines       15564    15625   +61     
=======================================
+ Hits        15395    15456   +61     
  Misses        169      169

Impacted Files	Coverage Δ
pennylane/templates/subroutines/__init__.py	`100.00% <100.00%> (ø)`
pennylane/templates/subroutines/kupccgsd.py	`100.00% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 06f2fe8...4458507. Read the comment docs.

pennylane/templates/subroutines/kupccgsd.py

soranjh · 2021-10-18T15:55:16Z

pennylane/templates/subroutines/kupccgsd.py

+           spin-orbitals included in the active space, and should be even.
+
+        #. The number of trainable parameters scales linearly with the number of layers as
+           :math:`2 k ||\vec{\theta}||`, where :math:`||\vec{\theta}||` is the total number of


Maybe use a symbol like n for representing the total number of excitation terms?

I have updated it. Does it look better now?

soranjh · 2021-10-18T15:55:23Z

pennylane/templates/subroutines/kupccgsd.py

+        Args:
+            k (int): Number of layers
+            n_wires (int): Number of qubits
+            delta_saz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``


Suggested change

delta_saz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``

delta_sz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``

soranjh · 2021-10-18T16:01:45Z

tests/templates/test_subroutines/test_kupccgsd.py

+                6,
+                6,
+                qml.math.array(
+                    [


Are we sure the reference state is the correct one?

Yes. I have computed them manually to cross-check.

soranjh · 2021-10-18T16:03:54Z

tests/templates/test_subroutines/test_kupccgsd.py

+        ],
+    )
+    def test_shape(self, k, n_wires, delta_sz, expected_shape):
+        """Test that the shape method returns the correct shape of the weights tensor"""


Suggested change

"""Test that the shape method returns the correct shape of the weights tensor"""

"""Test that the shape method returns the correct shape of the weights tensor."""

soranjh · 2021-10-18T16:06:32Z

tests/templates/test_subroutines/test_kupccgsd.py

+
+
+class TestInterfaces:
+    """Tests that the template is compatible with all interfaces, including the computation


We should change "Tests" to "Test" for consistency, everywhere.

I think so. I have made the suggested changes.

soranjh · 2021-10-18T16:45:34Z

tests/templates/test_subroutines/test_kupccgsd.py

+
+
+def circuit_decomposed(weights):
+    qml.BasisState(np.array([0, 0, 1, 1]), wires=[0, 1, 2, 3])


Any reason why the initial state is flipped?

Mainly because I am using the same wire conventions as qchem.excitations_to_wires. The state is flipped in the UCCSD template as well, which I suppose is related to wire indices in SingleExcitationUnitary and DoubleExcitationUnitary.

Does the test pass if we use np.array([1, 1, 0, 0])?

I just tested and it didn't pass. The expectation value Z0 returned for np.array([1, 1, 0, 0]) is opposite in sign in this case.

Good catch Soran. Yes, this is a known issue with these old excitation unitaries. It's in fact one of the reasons the tutorial on VQE in different spin sectors had to be changed.

Another reason to revisit these old excitations!

soranjh · 2021-10-18T16:46:37Z

pennylane/templates/subroutines/kupccgsd.py

+from pennylane.operation import Operation, AnyWires
+
+
+def generalized_singles(wires, delta_sz):


Is this function tested?

Thanks for pointing it out. I have now added a test for this and generalized_pair_doubles.

soranjh · 2021-10-18T16:46:45Z

pennylane/templates/subroutines/kupccgsd.py

+    return gen_singles_wires
+
+
+def generalized_pair_doubles(wires):


Is this function tested?

Thanks for pointing it out. It is now. Look at my comment above.

What is "generalized" here and in generalized_singles?

soranjh

Looks good to me, left few suggestions and questions.

…ndsurrender/pennylane into add-kupccgsd-template

pennylane/templates/subroutines/kupccgsd.py

Co-authored-by: Jay Soni <jbsoni@uwaterloo.ca>

ixfoduap

Looks very good overall, just minor suggestions

ixfoduap · 2021-10-15T18:29:22Z

doc/introduction/templates.rst

@@ -178,6 +178,11 @@ Other useful templates which do not belong to the previous categories can be fou
    :description: UCCSD
    :figure: ../_static/templates/subroutines/uccsd.png  

+.. customgalleryitem::
+    :link: ../code/api/pennylane.templates.subroutines.kUpCCGSD.html
+    :description: k-UpCCGSD


Small nitpick: Please change the kets with dots |.> on the left of the image with just dots
|0>
|0>
.
.
|1>
|1>

I have updated the image.

ixfoduap · 2021-10-19T18:41:01Z

pennylane/templates/subroutines/kupccgsd.py

+
+
+class kUpCCGSD(Operation):
+    r"""Implements the k-Unitary pair Coupled-Cluster Generalized Singles and Doubles (k-UpCCGSD) ansatz.


Suggested change

r"""Implements the k-Unitary pair Coupled-Cluster Generalized Singles and Doubles (k-UpCCGSD) ansatz.

r"""Implements the k-Unitary Pair Coupled-Cluster Generalized Singles and Doubles (k-UpCCGSD) ansatz.

I have corrected it.

ixfoduap · 2021-10-19T18:43:25Z

pennylane/templates/subroutines/kupccgsd.py

+    r"""Implements the k-Unitary pair Coupled-Cluster Generalized Singles and Doubles (k-UpCCGSD) ansatz.
+
+    The k-UpCCGSD ansatz calls the :func:`~.SingleExcitationUnitary` and :func:`~.DoubleExcitationUnitary`
+    templates to exponentiate the product of :math:`k` generalized singles and pair coupled-cluster doubles


It may be good to explain what we mean by "generalized" here. @obliviateandsurrender You can use a more compact version of the message you sent me on slack

NOTE: we should re-visit these excitation unitaries to understand if we can improve their implementation. Right now it is very slow

I have elucidated it in the text. Does it look sufficient?

ixfoduap · 2021-10-19T18:45:04Z

pennylane/templates/subroutines/kupccgsd.py

+            single and pair double excitation terms.
+        wires (Iterable): wires that the template acts on
+        k (int): Number of times UpCCGSD unitary is repeated.
+        delta_saz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``


Suggested change

delta_saz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``

delta_saz (int): Specifies the selection rule ``sz[p] - sz[r] = delta_sz``

Updated it.

ixfoduap · 2021-10-19T18:46:09Z

pennylane/templates/subroutines/kupccgsd.py

+           spin-orbitals included in the active space, and should be even.
+
+        #. The number of trainable parameters scales linearly with the number of layers as
+           :math:`2 k n`, where :math:`n = ||\vec{\theta}||` is the total number of


Suggested change

:math:`2 k n`, where :math:`n = ||\vec{\theta}||` is the total number of

:math:`2 k n`, where :math:`n` is the total number of

Updated it.

ixfoduap · 2021-10-19T18:46:39Z

pennylane/templates/subroutines/kupccgsd.py

+            H, qubits = qml.qchem.molecular_hamiltonian(symbols, coordinates)
+
+            # Define the Hartree-Fock state
+            delta_sz = 0


Suggested change

delta_sz = 0

Updated this and its subsequent usages.

pennylane/templates/subroutines/kupccgsd.py

ixfoduap · 2021-10-19T18:50:13Z

pennylane/templates/subroutines/kupccgsd.py

+    return gen_singles_wires
+
+
+def generalized_pair_doubles(wires):


What is "generalized" here and in generalized_singles?

ixfoduap · 2021-10-19T20:34:04Z

tests/templates/test_subroutines/test_kupccgsd.py

+
+
+def circuit_decomposed(weights):
+    qml.BasisState(np.array([0, 0, 1, 1]), wires=[0, 1, 2, 3])


Good catch Soran. Yes, this is a known issue with these old excitation unitaries. It's in fact one of the reasons the tutorial on VQE in different spin sectors had to be changed.

Another reason to revisit these old excitations!

obliviateandsurrender · 2021-10-21T13:47:24Z

@soranjh @Jaybsoni, are there any additional queries that I need to address?

Jaybsoni · 2021-10-21T13:59:19Z

@soranjh @Jaybsoni, are there any additional queries that I need to address?

Looks good on my side !

obliviateandsurrender added 2 commits October 12, 2021 19:31

add kupccgsd template

86d6c39

remove print statement

e34714e

obliviateandsurrender added 2 commits October 13, 2021 11:18

Update changelog-dev.md

0f04d65

Merge branch 'master' into add-kupccgsd-template

9200dd0

ixfoduap requested review from ixfoduap, soranjh and Jaybsoni October 14, 2021 16:03

josh146 mentioned this pull request Oct 16, 2021

Add support for easily creating expansion functions given a device #1760

Merged

Jaybsoni reviewed Oct 18, 2021

View reviewed changes

pennylane/templates/subroutines/kupccgsd.py Outdated Show resolved Hide resolved

soranjh reviewed Oct 18, 2021

View reviewed changes

soranjh requested changes Oct 18, 2021

View reviewed changes

obliviateandsurrender added 3 commits October 19, 2021 14:31

add test for excitation wire indices

c98c939

Merge branch 'add-kupccgsd-template' of https://github.com/obliviatea…

73811de

…ndsurrender/pennylane into add-kupccgsd-template

Merge branch 'master' into add-kupccgsd-template

d615628

Jaybsoni reviewed Oct 19, 2021

View reviewed changes

pennylane/templates/subroutines/kupccgsd.py Show resolved Hide resolved

Jaybsoni reviewed Oct 19, 2021

View reviewed changes

pennylane/templates/subroutines/kupccgsd.py Outdated Show resolved Hide resolved

obliviateandsurrender and others added 3 commits October 19, 2021 23:12

update generalized_singles

37d4b1d

Co-authored-by: Jay Soni <jbsoni@uwaterloo.ca>

Merge branch 'master' into add-kupccgsd-template

7135192

update generalized_singles

89b1bfe

ixfoduap requested changes Oct 19, 2021

View reviewed changes

obliviateandsurrender added 2 commits October 20, 2021 23:10

update kupccgsd docs and image

78f3c92

Merge branch 'master' into add-kupccgsd-template

387f069

ixfoduap approved these changes Oct 21, 2021

View reviewed changes

Jaybsoni self-requested a review October 21, 2021 13:59

Jaybsoni approved these changes Oct 21, 2021

View reviewed changes

soranjh approved these changes Oct 21, 2021

View reviewed changes

obliviateandsurrender added 2 commits October 21, 2021 21:24

Merge branch 'master' into add-kupccgsd-template

d6314a8

Merge branch 'master' into add-kupccgsd-template

4458507

obliviateandsurrender merged commit 265ec33 into PennyLaneAI:master Oct 21, 2021

	delta_saz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``
	delta_sz (int): Specifies the selection rules ``sz[p] - sz[r] = delta_sz``

	"""Test that the shape method returns the correct shape of the weights tensor"""
	"""Test that the shape method returns the correct shape of the weights tensor."""



		class TestInterfaces:
		"""Tests that the template is compatible with all interfaces, including the computation



		def circuit_decomposed(weights):
		qml.BasisState(np.array([0, 0, 1, 1]), wires=[0, 1, 2, 3])

		from pennylane.operation import Operation, AnyWires


		def generalized_singles(wires, delta_sz):

		return gen_singles_wires


		def generalized_pair_doubles(wires):



		class kUpCCGSD(Operation):
		r"""Implements the k-Unitary pair Coupled-Cluster Generalized Singles and Doubles (k-UpCCGSD) ansatz.

	:math:`2 k n`, where :math:`n = \|\|\vec{\theta}\|\|` is the total number of
	:math:`2 k n`, where :math:`n` is the total number of

Add k-UpCCGSD Template #1743

Add k-UpCCGSD Template #1743

Conversation

obliviateandsurrender commented Oct 12, 2021

codecov bot commented Oct 12, 2021 • edited Loading

Codecov Report

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

soranjh Oct 18, 2021 • edited Loading

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

soranjh left a comment

Choose a reason for hiding this comment

ixfoduap 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

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

obliviateandsurrender commented Oct 21, 2021

Jaybsoni commented Oct 21, 2021

codecov bot commented Oct 12, 2021 •

edited

Loading

soranjh Oct 18, 2021 •

edited

Loading