Add Quantum Counting algorithm implementation by axif0 · Pull Request #176 · amazon-braket/amazon-braket-algorithm-library

axif0 · 2026-02-15T19:42:26Z

Issue #, if available:
#1193

Description of changes:
Add the Quantum Counting algorithm (Brassard et al., 1998) — uses QPE on the Grover operator to estimate the number of marked items in an unstructured search space. Includes source module, unit tests, tutorial notebook, and README entry.
Testing done:

Merge Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.

General

I have read the CONTRIBUTING doc
I used the commit message format described in CONTRIBUTING
I have updated any necessary documentation, including READMEs and API docs (if appropriate)

Tests

I have added tests that prove my fix is effective or that my feature works (if appropriate)
I have checked that my tests are not configured for a specific region or account (if appropriate)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

sesmart · 2026-02-17T15:11:15Z

Hi @axif0, thanks for the contribution! Updating from main should fix the readthedocs issue, but you likely need to run tox -e linters as well. Thanks!

codecov · 2026-02-18T14:06:26Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (3c92423) to head (2b6d28f).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##              main      #176    +/-   ##
==========================================
  Coverage   100.00%   100.00%            
==========================================
  Files           30        32     +2     
  Lines         1220      1358   +138     
  Branches       156       174    +18     
==========================================
+ Hits          1220      1358   +138

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

…surement counts.

sesmart

Thanks for the contribution! Overall the notebook looks very nice, good edge cases, left some concerns below.

sesmart · 2026-02-20T14:57:03Z

+    controlled_matrix = np.kron(p0, id_matrix) + np.kron(p1, grover_unitary)
+
+    targets = [control] + list(target_qubits)
+    circ.unitary(matrix=controlled_matrix, targets=targets)


While this does work, unfortunately, this is not quite what we are looking for 🥲 - it would probably be okay for a black box oracle, but here we can actually build the circuits, though this can certainly be valuable for building an intuition and confirming our logic.

Can you maybe grab some of the circuits from the Grover's file?

The Grover operator is now built entirely from circuit primitives: build_oracle() for the phase oracle and amplify() for diffusion, both from grovers_search.py. The unitary is only extracted (to_unitary()) for the controlled-G operation in QPE, which is necessary since implementing arbitrary controlled multi-gate circuits natively is non-trivial.

I made an error in suggesting to remove the run function -> the more common pattern here is to have a run_algorithm function, so if you could restore that, it would be great. My apologies.

I think there are maybe two options here, and it should close it out. We would like for there to be an actual circuit, and I agree that while the controlled unitary is non trivial, here there might be a reasonable way.

(1) For this particular U, I think you might be able to leverage the nature of the oracle a bit more. You could drop the amplify function, and then just call the oracle for "1" + the relevant states. I.e. in the diffussion operator "00" -> "100", and in the oracle, "xx" -> "1xx". Then you just need to implement the controlled H, with a Z on the ancilla to correct the phase. Does this make sense? This of course leads to an exponentially deep circuit, but that's a known issue with QPE.

(2) If this doesn't work, you can call qiskit-braket-provider from the run function before execution. If possible, we would like to limit this to 1- and 2q unitaries, which have standard and near-optimal solutions.

I've updated the PR to implement Option 1 (building the controlled-Grover operator from gate-level primitives). Here’s a summary of the changes:

Gate-Level Controlled Grover: I replaced the to_unitary() matrix approach with a circuit-based implementation.

Made the oracle controlled by prepending "1" to the marked states (adding the QPE qubit as an extra control on the MCZ).

Made the diffusion operator controlled using controlled-Hadamard gates (decomposed via S·H·T·CX·T†·H·S†) and a controlled zero-oracle.

Restored run_quantum_counting: Added it back to match the library's run_algorithm pattern and exported it in init.py.

Updated Tests: Updated to use the new gate-level circuits and added tests for
run_quantum_counting. All tests and linting pass.

Updated Notebook: Replaced direct device.run() calls with run_quantum_counting, regenerated the circuit diagrams to show the native gates, and updated the note to mention that the circuit is now actually more compatible with QPU execution."

Very nice! Looks good, am just running the tests.

…ng circuits instead of matrices.

yitchen-tim · 2026-02-24T20:09:16Z

+    Returns:
+        Circuit: The complete quantum counting circuit with result types.
+    """
+    counting_circ.add_circuit(


Avoid mutating circuits by pattern like

new_circuit = Circuit().add_circuit(counting_circ) ...

yitchen-tim · 2026-02-24T20:11:07Z

@sesmart I resolved the most of your comments that are solved. There is one that needs your eye on.

@axif0 I added a comment about avoiding side effect.

…ng a new circuit.

axif0 · 2026-02-26T15:33:04Z

@sesmart, @yitchen-tim hello, Is there anything else for improvement?

sesmart · 2026-02-26T16:33:46Z

+    controlled_matrix = np.kron(p0, id_matrix) + np.kron(p1, grover_unitary)
+
+    targets = [control] + list(target_qubits)
+    circ.unitary(matrix=controlled_matrix, targets=targets)


I made an error in suggesting to remove the run function -> the more common pattern here is to have a run_algorithm function, so if you could restore that, it would be great. My apologies.

I think there are maybe two options here, and it should close it out. We would like for there to be an actual circuit, and I agree that while the controlled unitary is non trivial, here there might be a reasonable way.

(1) For this particular U, I think you might be able to leverage the nature of the oracle a bit more. You could drop the amplify function, and then just call the oracle for "1" + the relevant states. I.e. in the diffussion operator "00" -> "100", and in the oracle, "xx" -> "1xx". Then you just need to implement the controlled H, with a Z on the ancilla to correct the phase. Does this make sense? This of course leads to an exponentially deep circuit, but that's a known issue with QPE.

(2) If this doesn't work, you can call qiskit-braket-provider from the run function before execution. If possible, we would like to limit this to 1- and 2q unitaries, which have standard and near-optimal solutions.

…nting and add `run_quantum_counting` function.

sesmart · 2026-02-26T22:31:50Z

+
+    # Second set of controlled-H gates on search qubits
+    for sq in search_qubits:
+        _add_controlled_h(circ, control, sq)


Lastly, I think you need a Z gate here - grover is 2 ss* - I, but I believe the oracle here is I - 2ss*. Might have just missed it.

… inherent -1 global phase, removing the need for post-processing phase adjustments.

axif0 · 2026-02-27T17:12:17Z

@sesmart, could you please share any suggestions? I’ve pushed the latest updates.

sesmart · 2026-02-27T17:38:55Z

Looks good to go, thanks @axif0 !

sesmart · 2026-02-27T17:51:49Z

@axif0, can you please comment on one of the algorithms issues in the BDK? It needs to be assigned but you have to comment first.

axif0 added 2 commits February 12, 2026 20:23

feat: Add Quantum Counting algorithm implementation

5ec1480

Add cell IDs, clear execution counts, and update notebook format

423bd44

axif0 requested a review from a team as a code owner February 15, 2026 19:42

axif0 changed the title ~~Counting~~ Add Quantum Counting algorithm implementation Feb 15, 2026

Refactor notebook formatting and improve code readability

c6fa3a9

axif0 and others added 2 commits February 18, 2026 23:51

test: add test case for get_quantum_counting_results with empty mea…

c8077f6

…surement counts.

Merge branch 'main' into counting

4a5b40c

sesmart reviewed Feb 20, 2026

View reviewed changes

axif0 and others added 3 commits February 21, 2026 02:04

refactor: reimplement Quantum Counting oracle and Grover operator usi…

ec74cb2

…ng circuits instead of matrices.

Merge branch 'main' into counting

5258311

Merge branch 'main' into counting

d69be1a

yitchen-tim reviewed Feb 24, 2026

View reviewed changes

prevent in-place modification of the input counting_circ by composi…

d837566

…ng a new circuit.

axif0 requested review from sesmart and yitchen-tim February 26, 2026 15:33

sesmart requested changes Feb 26, 2026

View reviewed changes

feat: Implement gate-level controlled Grover operator for Quantum Cou…

5b05e3f

…nting and add `run_quantum_counting` function.

axif0 requested a review from sesmart February 26, 2026 22:01

sesmart reviewed Feb 26, 2026

View reviewed changes

fix: Apply Z gate in the controlled diffusion operator to correct the…

2b6d28f

… inherent -1 global phase, removing the need for post-processing phase adjustments.

axif0 requested a review from sesmart February 27, 2026 17:10

sesmart approved these changes Feb 27, 2026

View reviewed changes

sesmart merged commit 894ba33 into amazon-braket:main Feb 27, 2026
13 checks passed

Conversation

axif0 commented Feb 15, 2026

Merge Checklist

General

Tests

Uh oh!

sesmart commented Feb 17, 2026

Uh oh!

codecov Bot commented Feb 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

sesmart left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yitchen-tim commented Feb 24, 2026

Uh oh!

axif0 commented Feb 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

axif0 commented Feb 27, 2026

Uh oh!

sesmart commented Feb 27, 2026

Uh oh!

sesmart commented Feb 27, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov Bot commented Feb 18, 2026 •

edited

Loading