def test_pull_does_not_change_state(self, k, e, rounds):
egreedy = EpsilonGreedy(num_arms=k, epsilon=e)
for _ in range(rounds):
egreedy.pull()
self.assertEqual(0, sum(egreedy.pull_counts))
self.assertEqual(0, sum(egreedy.rewards))
def test_pull_does_not_change_state_after_any_updates(
self, k, e, rounds, updates):
egreedy = EpsilonGreedy(num_arms=k, epsilon=e)
for _ in range(updates):
egreedy.update(chosen_arm=np.random.randint(0, k),
reward=np.random.random())
pulls = sum(egreedy.pull_counts)
rewards = sum(egreedy.rewards)
for _ in range(rounds):
egreedy.pull()
self.assertEqual(pulls, sum(egreedy.pull_counts))
self.assertEqual(rewards, sum(egreedy.rewards))
def test_pull_returns_a_valid_arm(self, k, e, rounds):
egreedy = EpsilonGreedy(num_arms=k, epsilon=e)
for _ in range(rounds):
arm = egreedy.pull()
self.assertGreaterEqual(arm, 0)
self.assertLess(arm, k)