Handle crashing micro simulations (#85)

* Add first draft of simulation crash handling

* Extend handling of crashing simulations to adaptivity and corner cases

* Adapt format in crash handling to pep8

* Add tests for simulation crashes

* Adapt formatting of tests

* Update logger message for crashing simulation in the first run

Co-authored-by: Ishaan Desai <[email protected]>

* Restructure simulation crash handling and add global condition for exit

* Add first draft of simulation crash handling

* Extend handling of crashing simulations to adaptivity and corner cases

* Adapt format in crash handling to pep8

* Add tests for simulation crashes

* Adapt formatting of tests

* Add interpolation to crash handling

* Add Inverse Distance Weighting and improve crash handling

* Format with pre-commit

* Apply suggestions from code review

Co-authored-by: Ishaan Desai <[email protected]>

* Incoporate review into crash handling

* Base crash interpolation on macro parameter

* Format simulation crash handling

* Adapt error for no available neighbor

* Apply text improvement suggestions from code review

Co-authored-by: Ishaan Desai <[email protected]>

* Apply suggestions from crash handling review

* Adapt interpolation test

---------

Co-authored-by: Torben Schiz <[email protected]>
Co-authored-by: Ishaan Desai <[email protected]>

.github/workflows/run-unit-tests.yml

@@ -24,6 +24,14 @@ jobs:
           pip3 install --user .
           pip3 uninstall -y pyprecice
 
-      - name: Run unit tests
+      - name: Run micro_manager unit test
         working-directory: micro-manager/tests/unit
         run: python3 -m unittest test_micro_manager.py
+
+      - name: Run interpolation unit test
+        working-directory: micro-manager/tests/unit
+        run: python3 -m unittest test_interpolation.py
+
+      - name: Run micro simulation crash unit test
+        working-directory: micro-manager/tests/unit
+        run: python3 -m unittest test_micro_simulation_crash_handling.py

micro_manager/interpolation.py

@@ -0,0 +1,83 @@
+import numpy as np
+from sklearn.neighbors import NearestNeighbors
+
+
+class Interpolation:
+    def __init__(self, logger):
+
+        self._logger = logger
+
+    def get_nearest_neighbor_indices(
+        self,
+        coords: np.ndarray,
+        inter_point: np.ndarray,
+        k: int,
+    ) -> np.ndarray:
+        """
+        Get local indices of the k nearest neighbors of a point.
+
+        Parameters
+        ----------
+        coords : list
+            List of coordinates of all points.
+        inter_point : list | np.ndarray
+            Coordinates of the point for which the neighbors are to be found.
+        k : int
+            Number of neighbors to consider.
+
+        Returns
+        ------
+        neighbor_indices : np.ndarray
+            Local indices of the k nearest neighbors in all local points.
+        """
+        if len(coords) < k:
+            self._logger.info(
+                "Number of desired neighbors k = {} is larger than the number of available neighbors {}. Resetting k = {}.".format(
+                    k, len(coords), len(coords)
+                )
+            )
+            k = len(coords)
+        neighbors = NearestNeighbors(n_neighbors=k).fit(coords)
+
+        neighbor_indices = neighbors.kneighbors(
+            [inter_point], return_distance=False
+        ).flatten()
+
+        return neighbor_indices
+
+    def interpolate(self, neighbors: np.ndarray, point: np.ndarray, values):
+        """
+            Interpolate a value at a point using inverse distance weighting. (https://en.wikipedia.org/wiki/Inverse_distance_weighting)
+            .. math::
+                f(x) = (\sum_{i=1}^{n} \frac{f_i}{\Vert x_i - x \Vert^2}) / (\sum_{j=1}^{n} \frac{1}{\Vert x_j - x \Vert^2})
+
+        Parameters
+        ----------
+        neighbors : np.ndarray
+            Coordinates at which the values are known.
+        point : np.ndarray
+            Coordinates at which the value is to be interpolated.
+        values :
+            Values at the known coordinates.
+
+        Returns
+        -------
+        interpol_val / summed_weights :
+            Value at interpolation point.
+        """
+        interpol_val = 0
+        summed_weights = 0
+        # Iterate over all neighbors
+        for inx in range(len(neighbors)):
+            # Compute the squared norm of the difference between interpolation point and neighbor
+            norm = np.linalg.norm(np.array(neighbors[inx]) - np.array(point)) ** 2
+            # If interpolation point is already part of the data it is returned as the interpolation result
+            # This avoids division by zero
+            if norm < 1e-16:
+                return values[inx]
+            # Update interpolation value
+            interpol_val += values[inx] / norm
+            # Extend normalization factor
+            summed_weights += 1 / norm
+
+        return interpol_val / summed_weights