def test_select_action_after_one_learning_step(self):
state, action, reward = 1, 1, 10
ql = learn.QLearner(num_states=2, num_actions=2, learning_rate=0.5)
ql.learn(state, action, reward)
selected_action = ql.act(0)
self.assertEqual(action, selected_action)
def test_change_state_after_learning(self):
state = 1
action = 0
reward = 0
ql = learn.QLearner(num_states=3, num_actions=2)
ql.learn(state, action, reward)
self.assertEqual(ql.current_state, state)
def test_learn_with_1_state_action_q_value(self):
state = 0
action = 0
reward = 10
ql = learn.QLearner(num_states=1, num_actions=1)
ql.learn(state, action, reward)
q_value = ql.q_values.get(state, action)
self.assertEqual(q_value, 10)
def test_learn_with_learning_rate(self):
state = 0
action = 0
expected_q_value = 10
rewards = [10, 10]
ql = learn.QLearner(num_states=1, num_actions=1, learning_rate=0.5)
for reward in rewards:
ql.learn(state, action, reward)
q_value = ql.q_values.get(state, action)
self.assertEqual(q_value, expected_q_value)
def test_learn_two_rewards(self):
state = 0
action = 0
rewards = [5, 10]
expected_q_value = 15
ql = learn.QLearner(num_states=1, num_actions=1)
for reward in rewards:
ql.learn(state, action, reward)
q_value = ql.q_values.get(state, action)
self.assertEqual(q_value, expected_q_value)
def test_learn_with_several_state_action_q_values(self):
current_state = 2
next_state = 4
action = 3
reward = 10
ql = learn.QLearner(initial_state=current_state,
num_states=5,
num_actions=4)
ql.learn(next_state, action, reward)
q_value = ql.q_values.get(current_state, action)
self.assertEqual(q_value, 10)
def test_select_action_after_several_learning_steps(self):
ql = learn.QLearner(num_states=2,
num_actions=2,
learning_rate=0.5,
discount_factor=0.5)
steps = [(1, 1, 10), (0, 0, 100), (1, 1, 5)]
excepted_actions = [(0, 1), (1, 0)]
for state, action, reward in steps:
ql.learn(state, action, reward)
for state, excepted_action in excepted_actions:
action = ql.act(state)
self.assertEqual(action, excepted_action)
def test_default_discount_factor(self):
expected_discount_factor = 1
ql = learn.QLearner()
self.assertEqual(ql.discount_factor, expected_discount_factor)
def test_default_current_state(self):
expected_current_state = 0
ql = learn.QLearner()
self.assertEqual(ql.current_state, expected_current_state)
def test_initial_learning_rate(self):
expected_learning_rate = 0.8
ql = learn.QLearner(learning_rate=0.8)
self.assertEqual(ql.learning_rate, expected_learning_rate)
def test_initial_discount_factor(self):
expected_discount_factor = 0.8
ql = learn.QLearner(discount_factor=0.8)
self.assertEqual(ql.discount_factor, expected_discount_factor)
def test_initial_state(self):
expected_current_state = 2
ql = learn.QLearner(initial_state=2)
self.assertEqual(ql.current_state, expected_current_state)
def test_default_learning_rate(self):
expected_learning_rate = 1
ql = learn.QLearner()
self.assertEqual(ql.learning_rate, expected_learning_rate)
