def __init__(self,
state_space_parameters,
epsilon,
state=None,
qstore=None,
replay_dictionary=pd.DataFrame(columns=['net',
'accuracy_best_val',
'accuracy_last_val',
'accuracy_best_test',
'accuracy_last_test',
'ix_q_value_update',
'epsilon'])):
self.state_list = []
self.state_space_parameters = state_space_parameters
# Class that will expand states for us
self.enum = se.StateEnumerator(state_space_parameters)
self.stringutils = StateStringUtils(state_space_parameters)
self.model=self._build_model()
# Starting State
self.state = se.State('start', 0, 1, 0, 0, state_space_parameters.image_size, 0, 0) if not state else state
self.bucketed_state = self.enum.bucket_state(self.state)
# Cached Q-Values -- used for q learning update and transition
self.qstore = QValues() if not qstore else qstore
self.replay_dictionary = replay_dictionary
self.epsilon = epsilon # epsilon: parameter for epsilon greedy strategy
def __init__(self):
self.start_state = se.State('start', 0, 1, 0, 0, ssp.image_size, 0, 0,
0, 0)
self.q_path = 'needed_for_testing/q_values.csv'
self.se = se.StateEnumerator(ssp)
def __init__(self):
self.se = se.StateEnumerator(ssp)
def __init__(self, state_space_parameters):
self.image_size = state_space_parameters.image_size
self.output_number = state_space_parameters.output_states
self.enum = se.StateEnumerator(state_space_parameters)