Insertion noise model (quantumlib#4672)

This PR is part of quantumlib#4640. It adds the `InsertionNoiseModel`, which injects noise based on a user-defined map.

quantumlib#4671 is a prerequisite for this PR. The only files that need to be reviewed in this PR are:

- `cirq-core/cirq/devices/...`
    - `__init__.py`
    - `insertion_noise_model[_test].py`

cirq-core/cirq/devices/__init__.py

@@ -48,6 +48,10 @@
     draw_placements,
 )
 
+from cirq.devices.insertion_noise_model import (
+    InsertionNoiseModel,
+)
+
 from cirq.devices.noise_utils import (
     OpIdentifier,
     decay_constant_to_xeb_fidelity,

cirq-core/cirq/devices/insertion_noise_model.py

@@ -0,0 +1,68 @@
+# Copyright 2021 The Cirq Developers
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import dataclasses
+from typing import TYPE_CHECKING, Dict, List, Optional, Sequence
+
+from cirq import devices, ops
+from cirq.devices import noise_utils
+
+if TYPE_CHECKING:
+    import cirq
+
+
+@dataclasses.dataclass
+class InsertionNoiseModel(devices.NoiseModel):
+    """Simple base noise model for inserting operations.
+
+    Operations generated by this model for a given moment are all added into a
+    single "noise moment", which is added before or after the original moment
+    based on `prepend`.
+
+    Args:
+        ops_added: a map of gate types (and optionally, qubits they act on) to
+            operations that should be added.
+        prepend: whether to add the new moment before the current one.
+        require_physical_tag: whether to only apply noise to operations tagged
+            with PHYSICAL_GATE_TAG.
+    """
+
+    ops_added: Dict[noise_utils.OpIdentifier, 'cirq.Operation'] = dataclasses.field(
+        default_factory=dict
+    )
+    prepend: bool = False
+    require_physical_tag: bool = True
+
+    def noisy_moment(
+        self, moment: 'cirq.Moment', system_qubits: Sequence['cirq.Qid']
+    ) -> 'cirq.OP_TREE':
+        noise_ops: List['cirq.Operation'] = []
+        candidate_ops = [
+            op
+            for op in moment
+            if (not self.require_physical_tag) or noise_utils.PHYSICAL_GATE_TAG in op.tags
+        ]
+        for op in candidate_ops:
+            match_id: Optional[noise_utils.OpIdentifier] = None
+            candidate_ids = [op_id for op_id in self.ops_added if op in op_id]
+            for op_id in candidate_ids:
+                if match_id is None or op_id.is_proper_subtype_of(match_id):
+                    match_id = op_id
+            if match_id is not None:
+                noise_ops.append(self.ops_added[match_id])
+        if not noise_ops:
+            return [moment]
+        if self.prepend:
+            return [ops.Moment(noise_ops), moment]
+        return [moment, ops.Moment(noise_ops)]

cirq-core/cirq/devices/insertion_noise_model_test.py

@@ -0,0 +1,103 @@
+# Copyright 2021 The Cirq Developers
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import cirq
+from cirq.devices.insertion_noise_model import InsertionNoiseModel
+from cirq.devices.noise_utils import (
+    PHYSICAL_GATE_TAG,
+    OpIdentifier,
+)
+
+
+def test_insertion_noise():
+    q0, q1 = cirq.LineQubit.range(2)
+    op_id0 = OpIdentifier(cirq.XPowGate, q0)
+    op_id1 = OpIdentifier(cirq.ZPowGate, q1)
+    model = InsertionNoiseModel(
+        {op_id0: cirq.T(q0), op_id1: cirq.H(q1)}, require_physical_tag=False
+    )
+    assert not model.prepend
+
+    moment_0 = cirq.Moment(cirq.X(q0), cirq.X(q1))
+    assert model.noisy_moment(moment_0, system_qubits=[q0, q1]) == [
+        moment_0,
+        cirq.Moment(cirq.T(q0)),
+    ]
+
+    moment_1 = cirq.Moment(cirq.Z(q0), cirq.Z(q1))
+    assert model.noisy_moment(moment_1, system_qubits=[q0, q1]) == [
+        moment_1,
+        cirq.Moment(cirq.H(q1)),
+    ]
+
+    moment_2 = cirq.Moment(cirq.X(q0), cirq.Z(q1))
+    assert model.noisy_moment(moment_2, system_qubits=[q0, q1]) == [
+        moment_2,
+        cirq.Moment(cirq.T(q0), cirq.H(q1)),
+    ]
+
+    moment_3 = cirq.Moment(cirq.Z(q0), cirq.X(q1))
+    assert model.noisy_moment(moment_3, system_qubits=[q0, q1]) == [moment_3]
+
+
+def test_prepend():
+    q0, q1 = cirq.LineQubit.range(2)
+    op_id0 = OpIdentifier(cirq.XPowGate, q0)
+    op_id1 = OpIdentifier(cirq.ZPowGate, q1)
+    model = InsertionNoiseModel(
+        {op_id0: cirq.T(q0), op_id1: cirq.H(q1)}, prepend=True, require_physical_tag=False
+    )
+
+    moment_0 = cirq.Moment(cirq.X(q0), cirq.Z(q1))
+    assert model.noisy_moment(moment_0, system_qubits=[q0, q1]) == [
+        cirq.Moment(cirq.T(q0), cirq.H(q1)),
+        moment_0,
+    ]
+
+
+def test_require_physical_tag():
+    q0, q1 = cirq.LineQubit.range(2)
+    op_id0 = OpIdentifier(cirq.XPowGate, q0)
+    op_id1 = OpIdentifier(cirq.ZPowGate, q1)
+    model = InsertionNoiseModel({op_id0: cirq.T(q0), op_id1: cirq.H(q1)})
+    assert model.require_physical_tag
+
+    moment_0 = cirq.Moment(cirq.X(q0).with_tags(PHYSICAL_GATE_TAG), cirq.Z(q1))
+    assert model.noisy_moment(moment_0, system_qubits=[q0, q1]) == [
+        moment_0,
+        cirq.Moment(cirq.T(q0)),
+    ]
+
+
+def test_supertype_matching():
+    # Demonstrate that the model applies the closest matching type
+    # if multiple types match a given gate.
+    q0 = cirq.LineQubit(0)
+    op_id0 = OpIdentifier(cirq.Gate, q0)
+    op_id1 = OpIdentifier(cirq.XPowGate, q0)
+    model = InsertionNoiseModel(
+        {op_id0: cirq.T(q0), op_id1: cirq.S(q0)}, require_physical_tag=False
+    )
+
+    moment_0 = cirq.Moment(cirq.Rx(rads=1).on(q0))
+    assert model.noisy_moment(moment_0, system_qubits=[q0]) == [
+        moment_0,
+        cirq.Moment(cirq.S(q0)),
+    ]
+
+    moment_1 = cirq.Moment(cirq.Y(q0))
+    assert model.noisy_moment(moment_1, system_qubits=[q0]) == [
+        moment_1,
+        cirq.Moment(cirq.T(q0)),
+    ]