Simplify pixel log-likelihood.

ctapipe/image/pixel_likelihood.py

@@ -9,7 +9,7 @@
 At high signal this simplifies to a gaussian approximation.
 
 The full and gaussian approximations are implemented, in addition to a general purpose
-implementation, which tries to intellegently switch 
+implementation, which tries to intellegently switch
 between the two. Speed tests are below:
 
 neg_log_likelihood_approx(image, prediction, spe, ped)
@@ -129,7 +129,7 @@ def neg_log_likelihood_numeric(
 
     likelihood = epsilon
 
-    ns = np.arange(*poisson(np.max(prediction)).ppf(confidence),)
+    ns = np.arange(*poisson(np.max(prediction)).ppf(confidence))
 
     ns = ns[ns >= 0]
 
@@ -177,13 +177,15 @@ def neg_log_likelihood(
     approx_mask = prediction > prediction_safety
 
     neg_log_l = 0
-    neg_log_l += neg_log_likelihood_approx(
-        image[approx_mask], prediction[approx_mask], spe_width, pedestal,
-    )
+    if np.sum(approx_mask) > 0:
+        neg_log_l += neg_log_likelihood_approx(
+            image[approx_mask], prediction[approx_mask], spe_width, pedestal,
+        )
 
-    neg_log_l += neg_log_likelihood_numeric(
-        image[~approx_mask], prediction[~approx_mask], spe_width, pedestal,
-    )
+    if np.sum(~approx_mask) > 0:
+        neg_log_l += neg_log_likelihood_numeric(
+            image[~approx_mask], prediction[~approx_mask], spe_width, pedestal,
+        )
 
     return neg_log_l
 

ctapipe/image/tests/test_pixel_likelihood.py

@@ -1,5 +1,5 @@
 import numpy as np
-from ctapipe.image import poisson_likelihood_full, poisson_likelihood_gaussian
+from ctapipe.image import neg_log_likelihood, neg_log_likelihood_approx
 
 
 def test_full_likelihood():
@@ -11,30 +11,29 @@ def test_full_likelihood():
     spe = 0.5  # Single photo-electron width
     pedestal = 1  # width of the pedestal distribution
 
-    image_small = [0, 1, 2]
-    expectation_small = [1, 1, 1]
+    image_small = np.array([0, 1, 2])
+    expectation_small = np.array([1, 1, 1])
 
-    full_like_small = poisson_likelihood_full(
-        image_small, expectation_small, spe, pedestal
+    full_like_small = neg_log_likelihood(image_small, expectation_small, spe, pedestal)
+    exp_diff = full_like_small - np.sum(
+        np.asarray([2.75630505, 2.62168656, 3.39248449])
     )
-    exp_diff = full_like_small - np.asarray([2.75630505, 2.62168656, 3.39248449])
-    exp_diff = np.sum(np.abs(exp_diff))
+
     # Check against known values
-    print(exp_diff)
     assert exp_diff / np.sum(full_like_small) < 1e-4
 
-    image_large = [40, 50, 60]
-    expectation_large = [50, 50, 50]
-    full_like_large = poisson_likelihood_full(
-        image_large, expectation_large, spe, pedestal
-    )
+    image_large = np.array([40, 50, 60])
+    expectation_large = np.array([50, 50, 50])
+
+    full_like_large = neg_log_likelihood(image_large, expectation_large, spe, pedestal)
     # Check against known values
-    exp_diff = full_like_large - np.asarray([7.45489137, 5.99305388, 7.66226007])
-    exp_diff = np.sum(np.abs(exp_diff))
+    exp_diff = full_like_large - np.sum(
+        np.asarray([7.45489137, 5.99305388, 7.66226007])
+    )
 
     assert exp_diff / np.sum(full_like_large) < 1e-4
 
-    gaus_like_large = poisson_likelihood_gaussian(
+    gaus_like_large = neg_log_likelihood_approx(
         image_large, expectation_large, spe, pedestal
     )