Fix parameter order in VQE, if the ansatz is resized (#7479) (#7526)

* fix VQE param order if ansatz is resized

* add test, rm trailing comment

* fix construct circuit for params != circuit params

* fix test_max_evals_grouped

In this test, the ansatz was resized and thus the parameters ASCII-sorted. Now they are sorted by vector, which leads to a slightly different optimization (since the params are in a different order).

Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
(cherry picked from commit d0b26c2)

Co-authored-by: Julien Gacon <gaconju@gmail.com>

qiskit/algorithms/minimum_eigen_solvers/vqe.py

@@ -141,8 +141,6 @@ def __init__(
         self.expectation = expectation
         self._include_custom = include_custom
 
-        # set ansatz -- still supporting pre 0.18.0 sorting
-        self._ansatz_params = None
         self._ansatz = None
         self.ansatz = ansatz
 
@@ -186,7 +184,6 @@ def ansatz(self, ansatz: Optional[QuantumCircuit]):
             ansatz = RealAmplitudes()
 
         self._ansatz = ansatz
-        self._ansatz_params = list(ansatz.parameters)
 
     @property
     def gradient(self) -> Optional[Union[GradientBase, Callable]]:
@@ -278,7 +275,6 @@ def _check_operator_ansatz(self, operator: OperatorBase):
                 # try to set the number of qubits on the ansatz, if possible
                 try:
                     self.ansatz.num_qubits = operator.num_qubits
-                    self._ansatz_params = sorted(self.ansatz.parameters, key=lambda p: p.name)
                 except AttributeError as ex:
                     raise AlgorithmError(
                         "The number of qubits of the ansatz does not match the "
@@ -382,8 +378,7 @@ def construct_expectation(
         else:
             expectation = self.expectation
 
-        param_dict = dict(zip(self._ansatz_params, parameter))  # type: Dict
-        wave_function = self.ansatz.assign_parameters(param_dict)
+        wave_function = self.ansatz.assign_parameters(parameter)
 
         observable_meas = expectation.convert(StateFn(operator, is_measurement=True))
         ansatz_circuit_op = CircuitStateFn(wave_function)
@@ -525,8 +520,8 @@ def compute_minimum_eigenvalue(
         # optimization routine.
         if isinstance(self._gradient, GradientBase):
             gradient = self._gradient.gradient_wrapper(
-                ~StateFn(operator) @ StateFn(self._ansatz),
-                bind_params=self._ansatz_params,
+                ~StateFn(operator) @ StateFn(self.ansatz),
+                bind_params=list(self.ansatz.parameters),
                 backend=self._quantum_instance,
             )
         else:
@@ -564,7 +559,7 @@ def compute_minimum_eigenvalue(
 
         result = VQEResult()
         result.optimal_point = opt_result.x
-        result.optimal_parameters = dict(zip(self._ansatz_params, opt_result.x))
+        result.optimal_parameters = dict(zip(self.ansatz.parameters, opt_result.x))
         result.optimal_value = opt_result.fun
         result.cost_function_evals = opt_result.nfev
         result.optimizer_time = eval_time
@@ -615,14 +610,15 @@ def get_energy_evaluation(
         if num_parameters == 0:
             raise RuntimeError("The ansatz must be parameterized, but has 0 free parameters.")
 
+        ansatz_params = self.ansatz.parameters
         expect_op, expectation = self.construct_expectation(
-            self._ansatz_params, operator, return_expectation=True
+            ansatz_params, operator, return_expectation=True
         )
 
         def energy_evaluation(parameters):
             parameter_sets = np.reshape(parameters, (-1, num_parameters))
             # Create dict associating each parameter with the lists of parameterization values for it
-            param_bindings = dict(zip(self._ansatz_params, parameter_sets.transpose().tolist()))
+            param_bindings = dict(zip(ansatz_params, parameter_sets.transpose().tolist()))
 
             start_time = time()
             sampled_expect_op = self._circuit_sampler.convert(expect_op, params=param_bindings)

releasenotes/notes/fix-vqe-paramorder-if-ansatz-resized-14634a7efff7c74f.yaml

@@ -0,0 +1,10 @@
+---
+fixes:
+  - |
+    Fix a bug in :class:`~qiskit.algorithms.VQE` where the parameters of the ansatz were 
+    still explicitly ASCII-sorted by their name if the ansatz was resized. That led to a 
+    mismatching order of the optimized values in the ``optimal_point`` attribute of the result
+    object.
+
+    This bug did in particular occur if no ansatz was set by the user and the VQE chose 
+    a default with 11 or more free parameters.

test/python/algorithms/test_vqe.py

@@ -46,6 +46,7 @@
     X,
     Z,
 )
+from qiskit.quantum_info import Statevector
 from qiskit.transpiler import PassManager, PassManagerConfig
 from qiskit.transpiler.preset_passmanagers import level_1_pass_manager
 from qiskit.utils import QuantumInstance, algorithm_globals, has_aer
@@ -170,7 +171,7 @@ def test_missing_varform_params(self):
 
     @data(
         (SLSQP(maxiter=50), 5, 4),
-        (SPSA(maxiter=150), 3, 2),  # max_evals_grouped=n or =2 if n>2
+        (SPSA(maxiter=150), 2, 2),  # max_evals_grouped=n or =2 if n>2
     )
     @unpack
     def test_max_evals_grouped(self, optimizer, places, max_evals_grouped):
@@ -714,6 +715,21 @@ def test_2step_transpile(self):
         ]
         self.assertEqual([record.message for record in cm.records], expected)
 
+    def test_construct_eigenstate_from_optpoint(self):
+        """Test constructing the eigenstate from the optimal point, if the default ansatz is used."""
+
+        # use Hamiltonian yielding more than 11 parameters in the default ansatz
+        hamiltonian = Z ^ Z ^ Z
+        optimizer = SPSA(maxiter=1, learning_rate=0.01, perturbation=0.01)
+        quantum_instance = QuantumInstance(
+            backend=BasicAer.get_backend("statevector_simulator"), basis_gates=["u3", "cx"]
+        )
+        vqe = VQE(optimizer=optimizer, quantum_instance=quantum_instance)
+        result = vqe.compute_minimum_eigenvalue(hamiltonian)
+
+        optimal_circuit = vqe.ansatz.bind_parameters(result.optimal_point)
+        self.assertTrue(Statevector(result.eigenstate).equiv(optimal_circuit))
+
 
 if __name__ == "__main__":
     unittest.main()