Initialized tests for the new normal RV

Author: marvinpfoertner

src/probnum/backend/_core/_torch.py

@@ -34,7 +34,9 @@
 
 
 def asdtype(x) -> torch.dtype:
-    # Parse `x` with NumPy and convert `np.dtype`` into `torch.dtype`
+    if isinstance(x, torch.dtype):
+        return x
+
     return torch.as_tensor(
         np.empty(
             (),

src/probnum/randprocs/_random_process.py

@@ -5,8 +5,7 @@
 
 import numpy as np
 
-from probnum import randvars
-from probnum import utils as _utils
+from probnum import randvars, utils as _utils
 from probnum.typing import DTypeArgType, IntArgType, ShapeArgType
 
 _InputType = TypeVar("InputType")
@@ -68,7 +67,7 @@ def __repr__(self) -> str:
         )
 
     @abc.abstractmethod
-    def __call__(self, args: _InputType) -> randvars.RandomVariable[_OutputType]:
+    def __call__(self, args: _InputType) -> randvars.RandomVariable:
         """Evaluate the random process at a set of input arguments.
 
         Parameters

tests/test_randvars/test_normal/__init__.py

@@ -0,0 +1,23 @@
+"""Test cases defining random variables with a normal distribution."""
+
+from pytest_cases import case, parametrize
+
+from probnum import backend, randvars
+from probnum.problems.zoo.linalg import random_spd_matrix
+from probnum.typing import ScalarLike
+
+
+@case(tags=["univariate"])
+@parametrize("mean", (-1.0, 1))
+@parametrize("var", (3.0, 2))
+def case_univariate(mean: ScalarLike, var: ScalarLike) -> randvars.Normal:
+    return randvars.Normal(mean, var)
+
+
+@case(tags=["vectorvariate"])
+@parametrize("dim", [1, 2, 5, 10, 20])
+def case_vectorvariate(dim: int) -> randvars.Normal:
+    mean = backend.random.standard_normal(backend.random.seed(654 + dim), shape=(dim,))
+    cov = random_spd_matrix(backend.random.seed(846), dim)
+
+    return randvars.Normal(mean, cov)

tests/test_randvars/test_normal/cases.py

@@ -0,0 +1,16 @@
+"""Test properties of normal random variables."""
+import numpy as np
+import scipy.stats
+from pytest_cases import parametrize_with_cases
+
+from probnum import backend
+
+
+@parametrize_with_cases("rv", cases=".cases", has_tag=["univariate"])
+def test_entropy(rv):
+    scipy_entropy = scipy.stats.norm.entropy(
+        loc=backend.to_numpy(rv.mean),
+        scale=backend.to_numpy(rv.std),
+    )
+
+    np.testing.assert_allclose(backend.to_numpy(rv.entropy), scipy_entropy)