def test_should_update_an_arm_with_feedback(self):
arm_names = ['arm1', 'arm2']
arm = [EpsilonGreedyArm(name, n=1, reward_sum=1) for name in arm_names]
egreedy = EGreedy('my_bandit', arm, epsilon=0.1)
feedback = Feedback('arm1', reward=1)
egreedy.update(feedback)
assert egreedy.arms_dict['arm1'].reward_sum == 2 and egreedy.arms_dict[
'arm1'].n == 2
def test_should_run_for_several_iterations_and_sample_all_arms(self):
arm_names = ['arm1', 'arm2', 'arm3']
arm = [EpsilonGreedyArm(name, n=1, reward_sum=1) for name in arm_names]
egreedy = EGreedy('my_bandit', arm, epsilon=0.1)
num_steps = 100
sampled_arm_ids = set()
for _ in range(num_steps):
arm = egreedy.choose()
sampled_arm_ids.add(arm.arm_id)
reward = 1.0 if random.random() < 0.2 else 0.0
egreedy.update(Feedback(arm.arm_id, reward))
assert sampled_arm_ids == set(arm_names)
def test_should_reset_value_to_default(self):
arm_names = ['arm1', 'arm2', 'arm3']
arm = [EpsilonGreedyArm(name, n=1, reward_sum=1) for name in arm_names]
egreedy = EGreedy('my_bandit', arm, epsilon=0.1)
num_steps = 20
for _ in range(num_steps):
arm = egreedy.choose()
reward = 1.0 if random.random() < 0.2 else 0.0
egreedy.update(Feedback(arm_id=arm.arm_id, reward=reward))
egreedy.reset()
assert all([
arm.reward_sum == 0 and arm.n == 0.001
for arm in egreedy.arms_dict.values()
])