Add random_state argument to linalg testing functions for seeded rand…

…omness (#2459)

cirq/testing/lin_alg_utils.py

@@ -21,7 +21,10 @@
 from cirq import linalg, value
 
 
-def random_superposition(dim: int) -> np.ndarray:
+def random_superposition(dim: int,
+                         *,
+                         random_state: value.RANDOM_STATE_LIKE = None
+                        ) -> np.ndarray:
     """Returns a random unit-length vector from the uniform distribution.
 
     Args:
@@ -30,8 +33,10 @@ def random_superposition(dim: int) -> np.ndarray:
     Returns:
         The sampled unit-length vector.
     """
-    state_vector = np.random.randn(dim).astype(complex)
-    state_vector += 1j * np.random.randn(dim)
+    random_state = value.parse_random_state(random_state)
+
+    state_vector = random_state.randn(dim).astype(complex)
+    state_vector += 1j * random_state.randn(dim)
     state_vector /= np.linalg.norm(state_vector)
     return state_vector
 
@@ -59,7 +64,8 @@ def random_unitary(dim: int, *,
     return q * (d / abs(d))
 
 
-def random_orthogonal(dim: int) -> np.ndarray:
+def random_orthogonal(dim: int, *, random_state: value.RANDOM_STATE_LIKE = None
+                     ) -> np.ndarray:
     """Returns a random orthogonal matrix distributed with Haar measure.
 
     Args:
@@ -72,7 +78,9 @@ def random_orthogonal(dim: int) -> np.ndarray:
         'How to generate random matrices from the classical compact groups'
         http://arxiv.org/abs/math-ph/0609050
     """
-    m = np.random.randn(dim, dim)
+    random_state = value.parse_random_state(random_state)
+
+    m = random_state.randn(dim, dim)
     q, r = np.linalg.qr(m)
     d = np.diag(r)
     return q * (d / abs(d))
@@ -96,7 +104,10 @@ def random_special_unitary(dim: int,
     return r
 
 
-def random_special_orthogonal(dim: int) -> np.ndarray:
+def random_special_orthogonal(dim: int,
+                              *,
+                              random_state: value.RANDOM_STATE_LIKE = None
+                             ) -> np.ndarray:
     """Returns a random special orthogonal matrix distributed with Haar measure.
 
     Args:
@@ -105,7 +116,7 @@ def random_special_orthogonal(dim: int) -> np.ndarray:
     Returns:
       The sampled special orthogonal matrix.
     """
-    m = random_orthogonal(dim)
+    m = random_orthogonal(dim, random_state=random_state)
     if np.linalg.det(m) < 0:
         m[0, :] *= -1
     return m

cirq/testing/lin_alg_utils_test.py

@@ -35,6 +35,13 @@ def test_random_superposition(dim):
     assert np.isclose(np.linalg.norm(state), 1.0)
 
 
+def test_random_superposition_deterministic_given_seed():
+    state1 = random_superposition(10, random_state=1234)
+    state2 = random_superposition(10, random_state=1234)
+
+    np.testing.assert_equal(state1, state2)
+
+
 def test_random_unitary():
     u1 = random_unitary(2)
     u2 = random_unitary(2)
@@ -49,6 +56,14 @@ def test_random_orthogonal():
     assert is_orthogonal(o2)
     assert not np.allclose(o1, o2)
 
+
+def test_random_orthogonal_deterministic_given_seed():
+    o1 = random_orthogonal(2, random_state=1234)
+    o2 = random_orthogonal(2, random_state=1234)
+
+    np.testing.assert_equal(o1, o2)
+
+
 def test_random_special_unitary():
     u1 = random_special_unitary(2)
     u2 = random_special_unitary(2)
@@ -72,6 +87,14 @@ def test_random_special_orthgonal():
     assert is_special_orthogonal(o2)
     assert not np.allclose(o1, o2)
 
+
+def test_random_special_orthogonal_deterministic_given_seed():
+    o1 = random_special_orthogonal(2, random_state=1234)
+    o2 = random_special_orthogonal(2, random_state=1234)
+
+    np.testing.assert_equal(o1, o2)
+
+
 def test_assert_allclose_up_to_global_phase():
     assert_allclose_up_to_global_phase(
         np.array([[1]]),