Add Clifford Removal Function

[nabe98]

Enables Clifford vertex removal.

• For [1], cases corresponding to the following lemmas are implemented

  • Lemma 4.7
  • Lemma 4.8
  • Lemma 4.9
  • Lemma 4.11
  • Lemma 4.12

• Tests

  • Lemma 4.7
  • Lemma 4.8
  • Lemma 4.9
  • Lemma 4.11
  • Lemma 4.12

[1] https://arxiv.org/abs/2003.01664

closes: #6

[masa10-f]

@nabe98
Could you check your code with mypy and pyright before my review? (although I know it's under development)

[masa10-f]

Great job! I added a few comments, so please refer to them.

Also, could you move ZXGraphState class into a new file, like zxgraphstate.py?

[masa10-f]

@nabe98 Could you also turn on the format/type check ci?

[masato-fuk]

@nabe98

When I run the following code, after Cliffords are removed from the original graph, there seems no gflow. Do you have any idea about this?

# %%
import numpy as np
import networkx as nx
import fastflow as ff
from fastflow import gflow
from graphix_zx.common import Plane
from graphix_zx.euler import is_clifford_angle
from graphix_zx.flow import FlowLike
from graphix_zx.graphstate import BaseGraphState
from graphix_zx.zxgraphstate import ZXGraphState
from graphix_zx.circuit import MBQCCircuit, circuit2graph
from graphix_zx.visualizer import visualize


def random_gflow_circ(
    width: int,
    depth: int,
    rng: np.random.Generator = np.random.default_rng(),
    edge_p: float = 0.5,
) -> MBQCCircuit:
    angle_list = [0, np.pi / 3, 2 * np.pi / 3, np.pi]
    circ = MBQCCircuit(width)
    for d in range(depth):
        for j in range(width):
            circ.j(j, rng.choice(angle_list))
        if d < depth - 1:
            for j in range(width):
                if rng.random() < edge_p:
                    circ.cz(j, (j + 1) % width)
            num = rng.integers(0, width)
            if num > 0:
                target = set(rng.choice([i for i in range(width)], num))
                circ.phase_gadget(target, rng.choice(angle_list))

    return circ


def check_clifford_removal(graphstate: BaseGraphState) -> bool:
    no_clifford = True
    for node in (
        graphstate.physical_nodes - graphstate.output_nodes - graphstate.input_nodes
    ):
        angle = graphstate.meas_angles[node]
        no_clifford &= not is_clifford_angle(angle)
    return no_clifford


def gflow_wrapper(graphstate: BaseGraphState) -> FlowLike:
    graph = nx.Graph()
    graph.add_nodes_from(graphstate.physical_nodes)
    graph.add_edges_from(graphstate.physical_edges)

    planes = graphstate.meas_planes
    fastflow_planes = {}
    for node, plane in planes.items():
        if plane == Plane.XY:
            fastflow_planes[node] = ff.common.Plane.XY
        elif plane == Plane.YZ:
            fastflow_planes[node] = ff.common.Plane.YZ
        elif plane == Plane.XZ:
            fastflow_planes[node] = ff.common.Plane.XZ
        else:
            msg = f"No match {plane}"
            raise ValueError(msg)

    gflow_object = gflow.find(
        graph, graphstate.input_nodes, graphstate.output_nodes, fastflow_planes
    ).f

    return gflow_object


# %%

# generate circuit

# graph, flow = get_random_flow_graph(5, 5)
circ = random_gflow_circ(5, 5)
graph, flow = circuit2graph(circ)
# visualize(graph, save=True, filename="test.png")
gflow_result = gflow_wrapper(graph)
print(gflow_result)

# %%

# ZX process

zxgraph = ZXGraphState()
zxgraph.append(graph)

# %%
zxgraph.remove_cliffords()
print("No Clifford?", check_clifford_removal(zxgraph))
# visualize(zxgraph, save=True, filename="zxtest.png")
gflow_result = gflow_wrapper(zxgraph)
print(gflow_result)

[nabe98]

@masa10-f
Sorry for the delayed response! The issue was caused by some errors in the measurement actions. I have confirmed that the gflow is successfully saved when running the sample code above.

Since it relies on the fastflow module, I didn't include the test to confirm that it preserves gflow. Do you think it should be included?

[masa10-f]

@nabe98 Hi! I added a few comments on the code. Please check them.

[masa10-f]

@nabe98 Could you move tests for ZXGraphState into a new file because the original file becomes too big to maintain?

[masa10-f]

@nabe98 Could you check ruff? Afterwards, you can merge this branch! Well done!
I'd like to simply the tests for maintainability, but it might be addressed in another issue.

[nabe98]

@masa10-f Thanks for reviewing!
I completed the ruff check now.