Expose QAOA initial points (#355)

* Update docplex.py

Expose QAOA initial points

* [pre-commit.ci] auto fixes from pre-commit.com hooks

* Update classification.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

* Update classification.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

* Update classification.py

* Update pyriemann_qiskit/utils/docplex.py

Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

* Update pyriemann_qiskit/classification.py

Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

* Update pyriemann_qiskit/classification.py

Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Quentin Barthélemy <q.barthelemy@gmail.com>

Author: 3 people

pyriemann_qiskit/classification.py

@@ -637,6 +637,7 @@ def _get_docplex_optimizer_from_params_bag(
     classical_optimizer,
     create_mixer,
     n_reps,
+    qaoa_initial_points,
 ):
     if quantum:
         if create_mixer:
@@ -653,6 +654,7 @@ def _get_docplex_optimizer_from_params_bag(
                 quantum_instance=quantum_instance,
                 upper_bound=upper_bound,
                 optimizer=qaoa_optimizer,
+                initial_points=qaoa_initial_points,
             )
     else:
         logger._log(f"Using ClassicalOptimizer ({type(classical_optimizer).__name__})")
@@ -681,6 +683,8 @@ class QuanticMDM(QuanticClassifierBase):
         Add qaoa_optimizer parameter.
     .. versionchanged:: 0.4.0
         Add QAOA-CV optimization.
+    .. versionchanged:: 0.4.1
+        Add qaoa_initial_points parameter.
 
     Parameters
     ----------
@@ -724,6 +728,8 @@ class QuanticMDM(QuanticClassifierBase):
     n_reps : int, default=3
         The number of time the mixer and cost operator are repeated in the QAOA-CV
         circuit.
+    qaoa_initial_points : Tuple[int, int], default=[0.0, 0.0].
+        Starting parameters (beta and gamma) for the NaiveQAOAOptimizer.
 
     See Also
     --------
@@ -760,6 +766,7 @@ def __init__(
         qaoa_optimizer=SLSQP(),
         create_mixer=None,
         n_reps=3,
+        qaoa_initial_points=[0.0, 0.0],
     ):
         QuanticClassifierBase.__init__(
             self, quantum, q_account_token, verbose, shots, None, seed
@@ -771,6 +778,7 @@ def __init__(
         self.qaoa_optimizer = qaoa_optimizer
         self.create_mixer = create_mixer
         self.n_reps = n_reps
+        self.qaoa_initial_points = qaoa_initial_points
 
     @staticmethod
     def _override_predict_distance(mdm):
@@ -815,6 +823,7 @@ def _init_algo(self, n_features):
             self.classical_optimizer,
             self.create_mixer,
             self.n_reps,
+            self.qaoa_initial_points,
         )
         set_global_optimizer(self._optimizer)
         return classifier
@@ -1093,6 +1102,8 @@ class QuanticNCH(QuanticClassifierBase):
         Add qaoa_optimizer parameter.
     .. versionchanged:: 0.4.0
         Add QAOA-CV optimization.
+    .. versionchanged:: 0.4.1
+        Add the qaoa_initial_points parameter.
 
     Parameters
     ----------
@@ -1141,6 +1152,8 @@ class QuanticNCH(QuanticClassifierBase):
     n_reps : int, default=3
         The number of time the mixer and cost operator are repeated in the QAOA-CV
         circuit.
+    qaoa_initial_points : Tuple[int, int], default=[0.0, 0.0].
+        Starting parameters (beta and gamma) for the NaiveQAOAOptimizer.
 
     References
     ----------
@@ -1165,6 +1178,7 @@ def __init__(
         qaoa_optimizer=SLSQP(),
         create_mixer=None,
         n_reps=3,
+        qaoa_initial_points=[0.0, 0.0],
     ):
         QuanticClassifierBase.__init__(
             self, quantum, q_account_token, verbose, shots, None, seed
@@ -1179,6 +1193,7 @@ def __init__(
         self.qaoa_optimizer = qaoa_optimizer
         self.create_mixer = create_mixer
         self.n_reps = n_reps
+        self.qaoa_initial_points = qaoa_initial_points
 
     def _init_algo(self, n_features):
         self._log("Nearest Convex Hull Classifier initiating algorithm")
@@ -1200,6 +1215,7 @@ def _init_algo(self, n_features):
             self.classical_optimizer,
             self.create_mixer,
             self.n_reps,
+            self.qaoa_initial_points,
         )
 
         # sets the optimizer for the distance functions

pyriemann_qiskit/utils/docplex.py

@@ -459,6 +459,8 @@ class NaiveQAOAOptimizer(pyQiskitOptimizer):
     optimizer : SciPyOptimizer, default=SLSQP()
         An instance of a scipy optimizer to find the optimal weights for the
         parametric circuit (ansatz).
+    initial_points : Tuple[int, int], default=[0.0, 0.0].
+        Starting parameters (beta and gamma) for the QAOA.
 
     Notes
     -----
@@ -479,11 +481,18 @@ class NaiveQAOAOptimizer(pyQiskitOptimizer):
     pyQiskitOptimizer
     """
 
-    def __init__(self, upper_bound=7, quantum_instance=None, optimizer=SLSQP()):
+    def __init__(
+        self,
+        upper_bound=7,
+        quantum_instance=None,
+        optimizer=SLSQP(),
+        initial_points=[0.0, 0.0],
+    ):
         pyQiskitOptimizer.__init__(self)
         self.upper_bound = upper_bound
         self.quantum_instance = quantum_instance
         self.optimizer = optimizer
+        self.initial_points = initial_points
 
     def convert_spdmat(self, X):
         """Convert a SPD matrix
@@ -566,7 +575,7 @@ def _callback(_eval_count, _weights, value, _meta):
         qaoa_mes = QAOA(
             sampler=quantum_instance,
             optimizer=self.optimizer,
-            initial_point=[0.0, 0.0],
+            initial_point=self.initial_points,
             callback=_callback,
         )
         qaoa = MinimumEigenOptimizer(qaoa_mes)