Select: use petab_select method to set estimated parameters

pypesto/select/misc.py

@@ -99,6 +99,22 @@ def correct_x_guesses(
     should be corrected by replacing them with
     - their corresponding values in `row` if possible, else
     - their corresponding nominal values in the `petab_problem.parameter_df`.
+
+    Parameters
+    ----------
+    x_guesses:
+        The startpoints to correct.
+    model:
+        The startpoints will be corrected to match this model.
+    petab_problem:
+        The model's corresponding PEtab problem. If this is not provided,
+        it will be created from the `model`.
+    hierarchical:
+        Whether hierarchical optimization is used.
+
+    Returns
+    -------
+    The corrected startpoint guesses.
     """
     # TODO reconsider whether correcting is a good idea (`x_guess` is no longer
     # the latest MLE then). Similar todo exists in

pypesto/select/model_problem.py

@@ -165,16 +165,22 @@ def set_result(self, result: Result):
         self.model.set_criterion(Criterion.NLLH, float(self.best_start.fval))
         self.model.compute_criterion(criterion=self.criterion)
 
-        self.model.estimated_parameters = {
+        estimated_parameters = {
             id: float(value)
             for index, (id, value) in enumerate(
-                zip(
-                    self.pypesto_problem.x_names,
-                    self.best_start.x,
-                )
+                dict(
+                    zip(
+                        self.pypesto_problem.x_names,
+                        self.best_start.x,
+                    )
+                ).items()
             )
             if index in self.pypesto_problem.x_free_indices
         }
+        self.model.set_estimated_parameters(
+            estimated_parameters=estimated_parameters,
+            scaled=True,
+        )
 
         if self.postprocessor is not None:
             self.postprocessor(self)

setup.cfg

@@ -163,7 +163,7 @@ select =
     # Remove when vis is moved to PEtab Select version
     networkx >= 2.5.1
     # End remove
-    petab-select >= 0.1.8
+    petab-select >= 0.1.12
 test =
     pytest >= 5.4.3
     pytest-cov >= 2.10.0

test/select/test_select.py

@@ -24,12 +24,14 @@
 from pypesto.select import model_problem
 from pypesto.select.misc import correct_x_guesses
 
-# Options sent to `pypesto.optimize.optimize.minimize`, to reduce run time.
-minimize_options = {
-    'engine': pypesto.engine.MultiProcessEngine(),
-    'n_starts': 20,
-    'filename': None,
-    'progress_bar': False,
+model_problem_options = {
+    # Options sent to `pypesto.optimize.optimize.minimize`, to reduce run time.
+    'minimize_options': {
+        'engine': pypesto.engine.MultiProcessEngine(),
+        'n_starts': 20,
+        'filename': None,
+        'progress_bar': False,
+    }
 }
 # Tolerances for the differences between expected and test values.
 tolerances = {
@@ -137,7 +139,7 @@ def test_problem_select(pypesto_select_problem):
     for expected_result in expected_results:
         best_model, _ = pypesto_select_problem.select(
             criterion=criterion,
-            minimize_options=minimize_options,
+            model_problem_options=model_problem_options,
             predecessor_model=best_model,
             candidate_space=candidate_space,
         )
@@ -186,7 +188,7 @@ def test_problem_select_to_completion(pypesto_select_problem):
         criterion=Criterion.BIC,
         select_first_improvement=True,
         startpoint_latest_mle=True,
-        minimize_options=minimize_options,
+        model_problem_options=model_problem_options,
         candidate_space=candidate_space,
     )
 
@@ -258,7 +260,7 @@ def test_problem_multistart_select(pypesto_select_problem, initial_models):
         method=Method.FORWARD,
         criterion=criterion,
         predecessor_models=initial_models,
-        minimize_options=minimize_options,
+        model_problem_options=model_problem_options,
     )
 
     expected_best_model_subspace_id = 'M1_3'
@@ -336,7 +338,7 @@ def test_postprocessors(petab_select_problem):
     best_model_1, newly_calibrated_models_1 = pypesto_select_problem.select(
         method=Method.FORWARD,
         criterion=Criterion.AIC,
-        minimize_options=minimize_options,
+        model_problem_options=model_problem_options,
     )
 
     expected_newly_calibrated_models_subspace_ids = ['M1_0']
@@ -403,7 +405,7 @@ def test_vis(pypesto_select_problem):
     best_models = pypesto_select_problem.select_to_completion(
         method=Method.FORWARD,
         criterion=criterion,
-        minimize_options=minimize_options,
+        model_problem_options=model_problem_options,
     )
     labels = {
         model.get_hash(): model.model_subspace_id