Fix tests

replay_classification/sorted_spikes.py

@@ -87,6 +87,7 @@ def poisson_log_likelihood(is_spike, conditional_intensity=None,
     '''
     probability_no_spike = -conditional_intensity * time_bin_size
     is_spike = atleast_kd(is_spike, conditional_intensity.ndim)
-    eps = np.spacing(1)
-    return (np.log(conditional_intensity + eps) * is_spike +
+    conditional_intensity[
+        np.isclose(conditional_intensity, 0.0)] = np.spacing(1)
+    return (np.log(conditional_intensity) * is_spike +
             probability_no_spike)

tests/test_core.py

@@ -38,11 +38,11 @@ def test_normalize_to_probability():
 
 
 @mark.parametrize('data, exponent, expected', [
-    (np.arange(1, 9), 1, np.nanprod(np.arange(1, 9))),
-    (np.arange(1, 9), 2, np.nanprod(np.arange(1, 9) ** 2)),  # test kwarg
-    (np.arange(1, 9).reshape(2, 4), 2,  # test product along 1st dimension
-     np.nanprod(np.arange(1, 9).reshape(2, 4) ** 2, axis=0)),
-    (2, 2, 4),  # test single data point
+    (np.arange(1, 9), 1, 1),
+    (np.arange(1, 9), 2, 1),  # test kwarg
+    (np.array([[0.2, 0.4], [0.1, 0.2]]), 2,
+     np.array([np.exp(-0.15), 1])),
+    (2, 2, 1),  # test single data point
 ])
 def test_combined_likelihood(data, exponent, expected):
     def likelihood_function(x, exponent=1):

tests/test_sorted_spikes.py

@@ -1,34 +1,23 @@
 import numpy as np
 from pytest import mark
 
-from replay_classification.sorted_spikes import poisson_likelihood
+from replay_classification.sorted_spikes import poisson_log_likelihood
+
+CONDITIONAL_INTENSITY = np.atleast_2d(np.array([0.20, 0.50, 0.20, 0.25]))
+SPIKE_LOG_LIKELIHOOD = np.log(CONDITIONAL_INTENSITY) - CONDITIONAL_INTENSITY
 
 
 @mark.parametrize('is_spike, expected_likelihood', [
-    (np.zeros(3,), np.array([[5, 2, 5, 4],
-                             [5, 2, 5, 4],
-                             [5, 2, 5, 4],
-                             ])),
-    (np.array([0, 1, 0]), np.array([[5, 2, 5, 4],
-                                    [5 * np.log(0.2), 2 * np.log(0.5),
-                                     5 * np.log(0.2), 4 * np.log(0.25)],
-                                    [5, 2, 5, 4],
-                                    ])),
-    (np.ones(3,), np.array([[5 * np.log(0.2), 2 * np.log(0.5),
-                             5 * np.log(0.2), 4 * np.log(0.25)],
-                            [5 * np.log(0.2), 2 * np.log(0.5),
-                             5 * np.log(0.2), 4 * np.log(0.25)],
-                            [5 * np.log(0.2), 2 * np.log(0.5),
-                             5 * np.log(0.2), 4 * np.log(0.25)],
-                            ])),
+    (np.zeros(3,), -CONDITIONAL_INTENSITY),
+    (np.array([0, 1, 0]), np.concatenate(
+        (-CONDITIONAL_INTENSITY,
+         SPIKE_LOG_LIKELIHOOD,
+         -CONDITIONAL_INTENSITY))),
+    (np.ones(3,), SPIKE_LOG_LIKELIHOOD * np.ones((3, 1))),
 ])
 def test_poisson_likelihood_is_spike(is_spike, expected_likelihood):
-    conditional_intensity = np.array(
-        [[np.log(0.2), np.log(0.5), np.log(0.2), np.log(0.25)],
-         [np.log(0.2), np.log(0.5), np.log(0.2), np.log(0.25)],
-         [np.log(0.2), np.log(0.5), np.log(0.2), np.log(0.25)]
-         ])
-    likelihood = poisson_likelihood(
-        is_spike, conditional_intensity=conditional_intensity,
+    ci = CONDITIONAL_INTENSITY * np.ones((3, 1))
+    likelihood = poisson_log_likelihood(
+        is_spike, conditional_intensity=ci,
         time_bin_size=1)
     assert np.allclose(likelihood, expected_likelihood)