Linear function usability improvements

qiskit/circuit/library/generalized_gates/linear_function.py

@@ -14,9 +14,10 @@
 
 from __future__ import annotations
 import numpy as np
-from qiskit.circuit import QuantumCircuit, Gate
+from qiskit.circuit.quantumcircuit import QuantumCircuit, Gate
 from qiskit.circuit.exceptions import CircuitError
 from qiskit.synthesis.linear import check_invertible_binary_matrix
+from qiskit.circuit.library.generalized_gates.permutation import PermutationGate
 
 
 class LinearFunction(Gate):
@@ -61,42 +62,36 @@ class LinearFunction(Gate):
     `Online at umich.edu. <https://web.eecs.umich.edu/~imarkov/pubs/jour/qic08-cnot.pdf>`_
     """
 
-    def __init__(
-        self,
-        linear: list[list[int]] | np.ndarray | QuantumCircuit,
-        validate_input: bool | None = False,
-    ) -> None:
+    def __init__(self, linear, validate_input=False):
         """Create a new linear function.
 
         Args:
-            linear (list[list] or ndarray[bool] or QuantumCircuit):
-                either an n x n matrix, describing the linear function,
-                or a quantum circuit composed of linear gates only
-                (currently supported gates are CX and SWAP).
+            linear (list[list] or ndarray[bool] or QuantumCircuit or LinearFunction
+                or PermutationGate or Clifford): data from which a linear function
+                can be constructed. It can be either a nxn matrix (describing the
+                linear transformation), a permutation (which is a special case of
+                a linear function), another linear function, a clifford (when it
+                corresponds to a linear function), or a quantum circuit composed of
+                linear gates (CX and SWAP) and other objects described above, including
+                nested subcircuits.
 
             validate_input: if True, performs more expensive input validation checks,
                 such as checking that a given n x n matrix is invertible.
 
         Raises:
             CircuitError: if the input is invalid:
-                either a matrix is non {square, invertible},
-                or a quantum circuit contains non-linear gates.
-
+                either the input matrix is not square or not invertible,
+                or the input quantum circuit contains non-linear objects
+                (for example, a Hadamard gate, or a Clifford that does
+                not correspond to a linear function).
         """
-        if not isinstance(linear, (list, np.ndarray, QuantumCircuit)):
-            raise CircuitError(
-                "A linear function must be represented either by a list, "
-                "a numpy array, or a quantum circuit with linear gates."
-            )
 
-        if isinstance(linear, QuantumCircuit):
-            # The following function will raise a CircuitError if there are nonlinear gates.
-            original_circuit = linear
-            linear = _linear_quantum_circuit_to_mat(linear)
+        # pylint: disable=cyclic-import
+        from qiskit.quantum_info import Clifford
 
-        else:
-            original_circuit = None
+        original_circuit = None
 
+        if isinstance(linear, (list, np.ndarray)):
             # Normalize to numpy array (coercing entries to 0s and 1s)
             try:
                 linear = np.array(linear, dtype=bool, copy=True)
@@ -116,10 +111,100 @@ def __init__(
                         "A linear function must be represented by an invertible matrix."
                     )
 
+        elif isinstance(linear, QuantumCircuit):
+            # The following function will raise a CircuitError if there are nonlinear gates.
+            original_circuit = linear
+            linear = LinearFunction._circuit_to_mat(linear)
+
+        elif isinstance(linear, LinearFunction):
+            linear = linear.linear.copy()
+
+        elif isinstance(linear, PermutationGate):
+            linear = LinearFunction._permutation_to_mat(linear)
+
+        elif isinstance(linear, Clifford):
+            # The following function will raise a CircuitError if clifford does not correspond
+            # to a linear function.
+            linear = LinearFunction._clifford_to_mat(linear)
+
+        # Note: if we wanted, we could also try to construct a linear function from a
+        # general operator, by first attempting to convert it to clifford, and then to
+        # a linear function.
+
+        else:
+            raise CircuitError("A linear function cannot be successfully constructed.")
+
         super().__init__(
             name="linear_function", num_qubits=len(linear), params=[linear, original_circuit]
         )
 
+    @staticmethod
+    def _circuit_to_mat(qc: QuantumCircuit):
+        """This creates a nxn matrix corresponding to the given quantum circuit."""
+        nq = qc.num_qubits
+        mat = np.eye(nq, nq, dtype=bool)
+
+        for instruction in qc.data:
+            if instruction.operation.name in ("barrier", "delay"):
+                # can be ignored
+                continue
+            if instruction.operation.name == "cx":
+                # implemented directly
+                cb = qc.find_bit(instruction.qubits[0]).index
+                tb = qc.find_bit(instruction.qubits[1]).index
+                mat[tb, :] = (mat[tb, :]) ^ (mat[cb, :])
+                continue
+            if instruction.operation.name == "swap":
+                # implemented directly
+                cb = qc.find_bit(instruction.qubits[0]).index
+                tb = qc.find_bit(instruction.qubits[1]).index
+                mat[[cb, tb]] = mat[[tb, cb]]
+                continue
+
+            # In all other cases, we construct the linear function for the operation.
+            # and compose (multiply) linear matrices.
+
+            if getattr(instruction.operation, "definition", None) is not None:
+                other = LinearFunction(instruction.operation.definition)
+            else:
+                other = LinearFunction(instruction.operation)
+
+            positions = [qc.find_bit(q).index for q in instruction.qubits]
+            other = other.extend_with_identity(len(mat), positions)
+            mat = np.dot(other.linear.astype(int), mat.astype(int)) % 2
+            mat = mat.astype(bool)
+
+        return mat
+
+    @staticmethod
+    def _clifford_to_mat(cliff):
+        """This creates a nxn matrix corresponding to the given Clifford, when Clifford
+        can be converted to a linear function. This is possible when the clifford has
+        tableau of the form [[A, B], [C, D]], with B = C = 0 and D = A^{-1}^t, and zero
+        phase vector. In this case, the required matrix is A^t.
+        Raises an error otherwise.
+        """
+        # Note: since cliff is a valid Clifford, then the condition D = A^{-1}^t
+        # holds automatically once B = C = 0.
+        if cliff.phase.any() or cliff.destab_z.any() or cliff.stab_x.any():
+            raise CircuitError("The given clifford does not correspond to a linear function.")
+        return np.transpose(cliff.destab_x)
+
+    @staticmethod
+    def _permutation_to_mat(perm):
+        """This creates a nxn matrix from a given permutation gate."""
+        nq = len(perm.pattern)
+        mat = np.zeros((nq, nq), dtype=bool)
+        for i, j in enumerate(perm.pattern):
+            mat[i, j] = True
+        return mat
+
+    def __eq__(self, other):
+        """Check if two linear functions represent the same matrix."""
+        if not isinstance(other, LinearFunction):
+            return False
+        return (self.linear == other.linear).all()
+
     def validate_parameter(self, parameter):
         """Parameter validation"""
         return parameter
@@ -140,7 +225,7 @@ def synthesize(self):
 
     @property
     def linear(self):
-        """Returns the n x n matrix representing this linear function"""
+        """Returns the n x n matrix representing this linear function."""
         return self.params[0]
 
     @property
@@ -171,22 +256,47 @@ def permutation_pattern(self):
         locs = np.where(linear == 1)
         return locs[1]
 
+    def extend_with_identity(self, num_qubits: int, positions: list[int]) -> LinearFunction:
+        """Extend linear function to a linear function over nq qubits,
+        with identities on other subsystems.
 
-def _linear_quantum_circuit_to_mat(qc: QuantumCircuit):
-    """This creates a n x n matrix corresponding to the given linear quantum circuit."""
-    nq = qc.num_qubits
-    mat = np.eye(nq, nq, dtype=bool)
-
-    for instruction in qc.data:
-        if instruction.operation.name == "cx":
-            cb = qc.find_bit(instruction.qubits[0]).index
-            tb = qc.find_bit(instruction.qubits[1]).index
-            mat[tb, :] = (mat[tb, :]) ^ (mat[cb, :])
-        elif instruction.operation.name == "swap":
-            cb = qc.find_bit(instruction.qubits[0]).index
-            tb = qc.find_bit(instruction.qubits[1]).index
-            mat[[cb, tb]] = mat[[tb, cb]]
-        else:
-            raise CircuitError("A linear quantum circuit can include only CX and SWAP gates.")
+        Args:
+            num_qubits: number of qubits of the extended function.
 
-    return mat
+            positions: describes the positions of original qubits in the extended
+                function's qubits.
+
+        Returns:
+            LinearFunction: extended linear function.
+        """
+        extended_mat = np.eye(num_qubits, dtype=bool)
+
+        for i, pos in enumerate(positions):
+            extended_mat[positions, pos] = self.linear[:, i]
+
+        return LinearFunction(extended_mat)
+
+    def mat_str(self):
+        """Return string representation of the linear function
+        viewed as a matrix with 0/1 entries.
+        """
+        return str(self.linear.astype(int))
+
+    def function_str(self):
+        """Return string representation of the linear function
+        viewed as a linear transformation.
+        """
+        out = "("
+        mat = self.linear
+        for row in range(self.num_qubits):
+            first_entry = True
+            for col in range(self.num_qubits):
+                if mat[row, col]:
+                    if not first_entry:
+                        out += " + "
+                    out += "x_" + str(col)
+                    first_entry = False
+            if row != self.num_qubits - 1:
+                out += ", "
+        out += ")\n"
+        return out

releasenotes/notes/linear-functions-usability-45265f293a80a6e5.yaml

@@ -0,0 +1,28 @@
+---
+features:
+  - |
+    Allowing to construct a :class:`.LinearFunction` object from more general quantum circuits,
+    that may contain:
+
+     * Barriers (of type :class:`~qiskit.circuit.Barrier`) and delays (:class:`~qiskit.circuit.Delay`),
+       which are simply ignored
+     * Permutations (of type :class:`~qiskit.circuit.library.PermutationGate`)
+     * Other linear functions
+     * Cliffords (of type :class:`.Clifford`), when the Clifford represents a linear function
+       (and a ``CircuitError`` exception is raised if not)
+     * Nested quantum circuits of this form
+
+  - |
+    Added :meth:`.LinearFunction.__eq__` method. Two objects of type :class:`.LinearFunction`
+    are considered equal when their representations as binary invertible matrices are equal.
+  - |
+    Added :meth:`.LinearFunction.extend_with_identity` method, which allows to extend
+    a linear function over ``k`` qubits to a linear function over ``n >= k`` qubits,
+    specifying the new positions of the original qubits and padding with identities on the
+    remaining qubits.
+  - |
+    Added two methods for pretty-printing :class:`.LinearFunction` objects:
+    :meth:`.LinearFunction.mat_str`, which returns the string representation of the linear
+    function viewed as a matrix with 0/1 entries, and
+    :meth:`.LinearFunction.function_str`, which returns the string representation of the
+    linear function viewed as a linear transformation.