[EXTREME PARAMETERS] create an attribute undefined_parameters in...

[EXTREME PARAMETERS] create an attribute undefined_parameters in ExtremeParameters object, to avoid crash when a shape is < 0 for point while we want to display

experiment/meteo_france_SCM_study/main_visualize.py

@@ -19,7 +19,7 @@ def study_iterator(study_class, only_first_one=False, both_altitude=False, verbo
     if verbose:
         print('Loading studies....')
     for nb_day in nb_days:
-        for alti in AbstractStudy.ALTITUDES[::1]:
+        for alti in AbstractStudy.ALTITUDES[::-1]:
             if verbose:
                 print('alti: {}, nb_day: {}'.format(alti, nb_day))
             study = study_class(alti, nb_day) if is_safran_study else study_class(alti)
@@ -36,11 +36,12 @@ def extended_visualization():
     for study_class in SCM_EXTENDED_STUDIES[1:2]:
         for study in study_iterator(study_class, only_first_one=True):
             study_visualizer = StudyVisualizer(study)
-            study_visualizer.visualize_all_kde_graphs()
+            # study_visualizer.visualize_all_kde_graphs()
+            study_visualizer.visualize_experimental_law()
 
 
 def normal_visualization():
-    for study_class in SCM_STUDIES[:1]:
+    for study_class in SCM_STUDIES[1:2]:
         for study in study_iterator(study_class, only_first_one=True):
             study_visualizer = StudyVisualizer(study)
             # study_visualizer.visualize_independent_margin_fits(threshold=[None, 20, 40, 60][0])
@@ -62,6 +63,6 @@ def complete_analysis(only_first_one=False):
 
 
 if __name__ == '__main__':
-    # normal_visualization()
-    extended_visualization()
+    normal_visualization()
+    # extended_visualization()
     # complete_analysis()

experiment/meteo_france_SCM_study/safran/safran_visualizer.py

@@ -43,6 +43,10 @@ class StudyVisualizer(object):
     def dataset(self):
         return AbstractDataset(self.observations, self.coordinates)
 
+    def visualize_experimental_law(self):
+        plot_name = ' experimental law'
+        self.show_or_save_to_file(plot_name)
+
     def visualize_all_kde_graphs(self):
         massif_names = self.study.safran_massif_names
         nb_columns = 5

extreme_estimator/extreme_models/margin_model/margin_function/abstract_margin_function.py

@@ -121,6 +121,7 @@ class AbstractMarginFunction(object):
         grid = []
         for i, xi in enumerate(linspace):
             gev_param = self.get_gev_params(np.array([xi]))
+            assert not gev_param.has_undefined_parameters, 'This case needs to be handled during display'
             grid.append(gev_param.summary_dict)
         grid = {gev_param: [g[gev_param] for g in grid] for gev_param in GevParams.SUMMARY_NAMES}
         return grid, linspace

extreme_estimator/margin_fits/extreme_params.py

 from abc import ABC
+import numpy as np
 
 from extreme_estimator.margin_fits.abstract_params import AbstractParams
 
@@ -13,4 +14,8 @@ class ExtremeParams(AbstractParams, ABC):
         self.location = loc
         self.scale = scale
         self.shape = shape
-        assert self.scale > 0
+        # Scale cannot be negative
+        # (sometimes it happens, when we want to find a quantile for every point of a 2D map
+        # then it can happen that a corner point that was not used for fitting correspond to a negative scale,
+        # in the case we set all the parameters as equal to np.nan, and we will not display those points)
+        self.has_undefined_parameters = self.scale <= 0
\ No newline at end of file

extreme_estimator/margin_fits/gev/gev_params.py

 from extreme_estimator.extreme_models.utils import r
 from extreme_estimator.margin_fits.extreme_params import ExtremeParams
+import numpy as np
 
 
 class GevParams(ExtremeParams):
@@ -14,8 +15,14 @@ class GevParams(ExtremeParams):
         self.block_size = block_size
 
     def quantile(self, p) -> float:
-        return r.qgev(p, self.location, self.scale, self.shape)[0]
+        if self.has_undefined_parameters:
+            return np.nan
+        else:
+            return r.qgev(p, self.location, self.scale, self.shape)[0]
 
     @property
     def param_values(self):
-        return [self.location, self.scale, self.shape]
+        if self.has_undefined_parameters:
\ No newline at end of file
+            return [np.nan for _ in range(3)]
+        else:
+            return [self.location, self.scale, self.shape]
\ No newline at end of file

extreme_estimator/margin_fits/plot/create_shifted_cmap.py

@@ -24,6 +24,8 @@ def plot_extreme_param(ax, gev_param_name, values):
         midpoint = 1.0
     elif vmin > 0 and vmax > 0:
         midpoint = 0.0
+    else:
+        raise ValueError('Unexpected values: vmin={}, vmax={}'.format(vmin, vmax))
     cmap = [plt.cm.coolwarm, plt.cm.bwr, plt.cm.seismic][1]
     shifted_cmap = shiftedColorMap(cmap, midpoint=midpoint, name='shifted')
     norm = mpl.colors.Normalize(vmin=vmin, vmax=vmax)
@@ -47,10 +49,9 @@ def get_color_rbga_shifted(ax, gev_param_name, values):
 
 
 def imshow_shifted(ax, gev_param_name, values, x, y):
-    norm, shifted_cmap = plot_extreme_param(ax, gev_param_name, values)
+    masked_array = np.ma.masked_where(np.isnan(values), values)
+    norm, shifted_cmap = plot_extreme_param(ax, gev_param_name, masked_array)
     shifted_cmap.set_bad(color='white')
-
-    masked_array = values
     if gev_param_name != ExtremeParams.SHAPE:
         epsilon = 1e-2 * (np.max(values) - np.min(values))
         value = np.min(values)

test/test_extreme_estimator/test_margin_fits/test_gev/test_gev_params.py

@@ -14,6 +14,12 @@ class TestGevParams(unittest.TestCase):
         for quantile_name, p in gev_params.quantile_name_to_p.items():
             self.assertAlmostEqual(- 1 / np.log(p), quantile_dict[quantile_name])
 
+    def test_negative_scale(self):
+        gev_params = GevParams(loc=1.0, shape=1.0, scale=-1.0)
+        for p in [0.1, 0.5, 0.9]:
+            q = gev_params.quantile(p)
+            self.assertTrue(np.isnan(q))
+
 
 if __name__ == '__main__':
     unittest.main()