def test_evaluate_episode_delta(self):
MC = MonteCarlo()
deltas = []
for _ in range(100):
deltas.append(MC.evaluate_episode_delta(MC.generate_episode()))
self.assertTrue(-1 <= deltas[-1] <= 1)
self.assertTrue(any(d for d in deltas))
print('sample path value max deltas:', deltas)
def test_generate_episode(self):
MC = MonteCarlo()
paths = []
for _ in range(10):
states = MC.generate_episode()
self.assertLess(states[-1], hash(MC.board))
paths.append(states)
print('sample state paths:', paths)
def test_evaluate_episode(self):
MC = MonteCarlo()
for _ in range(10):
states = MC.generate_episode()
pre_visits = np.array([MC.visits[s] for s in [0] + states])
pre_values = np.array([MC.values[s] for s in [0] + states])
MC.evaluate_episode(states)
post_visits = np.array([MC.visits[s] for s in [0] + states])
post_values = np.array([MC.values[s] for s in [0] + states])
diff_values = pre_values - post_values
diff_visits = pre_visits - post_visits
print('sample path value deltas:', diff_values)
self.assertTrue(np.all(diff_visits) == 1)