Add/forward startpoint_kwargs in PetabImporter.create_problem (#1135

)

Simplifies combining PEtab startpoint sampling with checked sampling (i.e. resample in case of non-finite fvals/gradients), e.g.:

```python
problem = importer.create_problem(
    startpoint_kwargs=dict(check_fval=True, check_grad=True)
)
```

Closes #1119

doc/example/petab_import.ipynb

@@ -127,8 +127,7 @@
       "\n",
       "Model outputs:    ['pSTAT5A_rel', 'pSTAT5B_rel', 'rSTAT5A_rel'] \n",
       "\n",
-      "Model states:     ['STAT5A', 'STAT5B', 'pApB', 'pApA', 'pBpB', 'nucpApA', 'nucpApB', 'nucpBpB'] \n",
-      "\n"
+      "Model states:     ['STAT5A', 'STAT5B', 'pApB', 'pApA', 'pBpB', 'nucpApA', 'nucpApB', 'nucpBpB'] \n"
      ]
     }
    ],
@@ -407,6 +406,30 @@
     ")"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "source": [
+    "### Dealing with function evaluations at the initial point\n",
+    "\n",
+    "It is quite common in real applications, that the objective function is evaluable at every point in parameter space. Therefore, some local optimizations may fail directly at their initial point. Such results are usually not very informative and would be discarded. To directly discard such initial points, we can select a startpoint method that will resample starting points if the objective function (`check_fval`) or its gradient (`check_grad`) are non-finite:"
+   ],
+   "metadata": {
+    "collapsed": false
+   }
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "outputs": [],
+   "source": [
+    "problem = importer.create_problem(\n",
+    "    startpoint_kwargs=dict(check_fval=True, check_grad=True)\n",
+    ")"
+   ],
+   "metadata": {
+    "collapsed": false
+   }
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -517,9 +540,6 @@
    "cell_type": "code",
    "execution_count": 15,
    "metadata": {
-    "jupyter": {
-     "outputs_hidden": false
-    },
     "pycharm": {
      "name": "#%%\n"
     }

pypesto/petab/importer.py

@@ -674,6 +674,7 @@ def create_problem(
         objective: AmiciObjective = None,
         x_guesses: Optional[Iterable[float]] = None,
         problem_kwargs: Dict[str, Any] = None,
+        startpoint_kwargs: Dict[str, Any] = None,
         **kwargs,
     ) -> Problem:
         """Create a :class:`pypesto.Problem`.
@@ -688,6 +689,9 @@ def create_problem(
             optimization.
         problem_kwargs:
             Passed to the `pypesto.Problem` constructor.
+        startpoint_kwargs:
+            Keyword arguments forwarded to
+            :meth:`PetabImporter.create_startpoint_method`.
         **kwargs:
             Additional key word arguments passed on to the objective,
             if not provided.
@@ -741,6 +745,9 @@ def create_problem(
         if problem_kwargs is None:
             problem_kwargs = {}
 
+        if startpoint_kwargs is None:
+            startpoint_kwargs = {}
+
         prior = self.create_prior()
 
         if prior is not None:
@@ -762,7 +769,7 @@ def create_problem(
             x_scales=x_scales,
             x_priors_defs=prior,
             startpoint_method=self.create_startpoint_method(
-                x_ids=np.delete(x_ids, x_fixed_indices)
+                x_ids=np.delete(x_ids, x_fixed_indices), **startpoint_kwargs
             ),
             **problem_kwargs,
         )