Updated some unittests.

Author: weidler

angorapy/common/policies.py

@@ -2,7 +2,7 @@
 
 import abc
 import math
-from typing import Union, Tuple
+from typing import Union, Tuple, List
 
 import gym
 import numpy as np
@@ -341,7 +341,7 @@ def _entropy_from_params(self, stdevs: tf.Tensor):
 
         Input Shape: (B, A) or (B, S, A) for recurrent
 
-        Since the given r.v.'serialization are independent, the subadditivity property of entropy narrows down to an equality
+        Since the given r.v.'s are independent, the sub-additivity property of entropy narrows down to an equality
         of the joint entropy and the sum of marginal entropies.
         """
         entropy = .5 * tf.math.log(2 * math.pi * math.e * tf.pow(stdevs, 2))
@@ -373,7 +373,7 @@ def _approx_entropy_from_log(self, log_stdevs: tf.Tensor):
         return tf.reduce_sum(log_stdevs, axis=-1)
 
     @tf.function
-    def entropy(self, params: tf.Tensor):
+    def entropy(self, params: Tuple[tf.Tensor, tf.Tensor]):
         """Calculate the joint entropy of Gaussian random variables described by their log standard deviations.
 
         Input Shape: (B, A) or (B, S, A) for recurrent

tests/test_distributions.py

@@ -39,7 +39,7 @@ def test_gaussian_entropy(self):
         sig = tf.convert_to_tensor([[1.0, 1.0], [1.0, 5.0]], dtype=tf.float32)
 
         result_reference = np.sum(norm.entropy(loc=mu, scale=sig), axis=-1)
-        result_log = distro.entropy(np.log(sig)).numpy()
+        result_log = distro.entropy([mu, np.log(sig)]).numpy()
         result = distro._entropy_from_params(sig).numpy()
 
         self.assertTrue(np.allclose(result_reference, result), msg="Gaussian entropy returns wrong result")