def test_action(self):
rewards = [PoissonReward(mu) for mu in np.random.rand(10) + 1]
e = Environment(rewards)
with pytest.raises(AssertionError):
e.action(11)
r = e.action(0)
print(r)
assert type(r) in [float, int]
def test_poisson_rewards(self):
n = 10
rewards = []
mus = np.linspace(1, 5, n)
for mu in mus:
rewards.append(PoissonReward(mu))
e = Environment(rewards)
assert len(e) == n
er = e.expected_rewards()
for i, r in enumerate(er):
assert r == rewards[i].expected_reward()
def test_gaussian_rewards(self):
n = 10
rewards = []
ms = np.linspace(-1, 1, n)
vs = np.linspace(0.1, 3, n)
for m, v in zip(ms, vs):
rewards.append(GaussianReward(m, v))
e = Environment(rewards)
assert len(e) == n
er = e.expected_rewards()
for i, r in enumerate(er):
assert r == rewards[i].expected_reward()
def setUp(self):
super().setUp()
N = 5
self.n_rewards = N
self.env = Environment([GaussianReward() for _ in range(N)])
class PosseTestCase(TestCase):
def setUp(self):
super().setUp()
Nr = 5
self.n_rewards = Nr
self.env = Environment([GaussianReward() for _ in range(Nr)])
def test_smoke(self):
posse = Posse(self.env, GreedyBandit, n_bandits=20)
assert isinstance(posse, Posse)
def test_numbers(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
assert len(posse.bandits) == N_bandits
assert posse.n_actions_taken == 0
assert posse.len_env == len(self.env)
assert posse.n_rewards == self.n_rewards
def test_actions(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
N_actions = 2
posse.take_actions(N_actions)
assert posse.n_actions_taken == 2
assert len(posse.bandits[0].choice_history) == 2
def test_bandit_kwargs(self):
N_bandits = 20
eps = 0.1
posse = Posse(self.env, EpsGreedyBandit, n_bandits=N_bandits, eps=eps)
assert posse.bandits[0].eps == eps
def test_mean_var_reward(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
N_actions = 100
posse.take_actions(N_actions)
assert posse.mean_reward().shape == (100, )
assert posse.var_reward().shape == (100, )
def test_mean_best_choice_with_int(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
N_actions = 100
posse.take_actions(N_actions)
bc = self.env.expected_rewards().argmax()
assert posse.mean_best_choice(bc).shape == (100, )
assert posse.var_best_choice(bc).shape == (100, )
def test_mean_best_choice(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
N_actions = 100
posse.take_actions(N_actions)
bc = np.zeros(100)
assert posse.mean_best_choice(bc).shape == (100, )
assert posse.mean_best_choice(0).shape == (100, )
assert posse.mean_best_choice(bc.tolist()).shape == (100, )
with pytest.raises(TypeError):
posse.mean_best_choice(3.1415)
def test_var_best_choice(self):
N_bandits = 20
posse = Posse(self.env, GreedyBandit, n_bandits=N_bandits)
N_actions = 100
posse.take_actions(N_actions)
bc = np.zeros(100)
assert posse.var_best_choice(bc).shape == (100, )
assert posse.var_best_choice(0).shape == (100, )
assert posse.var_best_choice(bc.tolist()).shape == (100, )
with pytest.raises(TypeError):
posse.var_best_choice(3.1415)
def test_moments(self):
rewards = [PoissonReward(mu) for mu in [4] * 5]
e = Environment(rewards)
moments = e.moments()
assert np.all(moments == 4)