def init_table(self) -> dict:
"""Returns a dict with reward statistics for every combination of oao|ar"""
table = {}
a_length = self.environment.action_length
o_length = self.environment.observation_length
if o_length == 0:
table[''] = {}
for action_idx in range(pow(2, a_length)):
action = Utility.get_bitstring_from_decimal(
action_idx, a_length)
table[''][action] = {}
table[''][action][''] = Utility.init_heap(a_length)
else:
for last_observation_idx in range(pow(2, o_length)):
last_observation = Utility.get_bitstring_from_decimal(
last_observation_idx, o_length)
table[last_observation] = {}
for action_idx in range(pow(2, a_length)):
action = Utility.get_bitstring_from_decimal(
action_idx, a_length)
table[last_observation][action] = {}
for observation_idx in range(pow(2, o_length)):
observation = Utility.get_bitstring_from_decimal(
observation_idx, o_length)
table[last_observation][action][
observation] = Utility.init_heap(a_length)
return table
def init_double_heap(length: int):
"""initialize heap with 2 actions for pi2forward agent"""
reward_statistics = []
for action_one_idx in range(pow(2, length)):
action_one = Utility.get_bitstring_from_decimal(action_one_idx, length)
for action_two_idx in range(pow(2, length)):
action_two = Utility.get_bitstring_from_decimal(action_two_idx, length)
# (expected_reward, actions, number_of_times)
heapq.heappush(reward_statistics, (2, action_one + action_two, 0))
return reward_statistics