Make CrossvalMultipleRegularizations more effective for very large da…

…tasets (#37)

Signed-off-by: Matthias Kümmmerer <[email protected]>

pysaliency/baseline_utils.py

@@ -369,17 +369,18 @@ def __init__(self, stimuli, fixations, regularization_models: OrderedDict, cross
             verbose=False
         )
 
-        mean_area = np.mean([x[2]*x[3] for x in X_areas])
-        self.mean_area = mean_area
 
         self.X = fixations_to_scikit_learn(
             self.fixations,
             normalize=self.stimuli,
             keep_aspect=True, add_shape=False, add_fixation_number=True, verbose=False
         )
 
-        real_areas = [self.stimuli.sizes[n][0]*self.stimuli.sizes[n][1] for n in self.fixations.n]
-        areas_gold = [x[2]*x[3] for x in X_areas]
+        stimuli_sizes = np.array(self.stimuli.sizes)
+        real_areas = stimuli_sizes[self.fixations.n, 0] * stimuli_sizes[self.fixations.n, 1]
+        areas_gold = X_areas[:, 2] * X_areas[:, 3]
+        self.mean_area = np.mean(areas_gold)
+
         correction = np.log(areas_gold) - np.log(real_areas)
         self.regularization_log_likelihoods = []
 

tests/test_baseline_utils.py

@@ -156,4 +156,5 @@ def test_crossval_multiple_regularizations(stimuli, fixation_trains):
     log_regularizations = [0.1, 0.2]
 
     score = estimator.score(log_bandwidth, *log_regularizations)
-    assert isinstance(score, float)
+    assert isinstance(score, float)
+    np.testing.assert_allclose(score, -1.4673831679692528e-10)