def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
self.q = np.random.rand(self.num_states, self.num_actions)
self.returns = {(state, action):[] for state in range(self.num_states) for action in range(self.num_actions)}
self.epsilon = 1
self.tuple_state_agent_met=[]
def __init__(self, params):
Agent.__init__(self, params)
# The 3 following args will be given by the env.
self.n_x = None
self.n_y = None
self.states = None
self.X_memory = []
self.Y_memory = []
self.ia = MLPRegressor(warm_start=True,
max_iter=200,
early_stopping=False,
hidden_layer_sizes=(20, 10, 5),
learning_rate_init=1 * 10**-3,
activation='identity')
self.epsilon = 0.5
self.gamma = 0.8
self.alpha = 0.5
self.last_action = None
self.last_state = None
self.is_action_possible = None
self.n_max_action = params["max_action_per_episode"]
self.first_fit = True
self.key_taken = []
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
#strategie initiale (au depart l'agent ne connait pas les deplacements optimaux a effectuer)
#tableau de taille nombre d'etat * nombre d'actions possibles
self.Q = np.zeros((int(params['num_cells_grid1D']),2))
#Probabilite d'exploration
self.exploration = 0.05
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
self.epsilon = 1
self.n_key_max = 5
self.alpha = 0.4
self.gamma = 0.9999
self.key_taken = []
self.q = np.random.rand(self.num_states, self.num_actions, self.n_key_max)
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
#Nombre de lignes
self.l = int(params['ligne'])
#Nombre de colonnes
self.c = int(params['colonne'])
#strategie initiale (au depart l'agent ne connait pas les deplacements optimaux a effectuer)
#tableau de taille nombre d'etat * nombre d'actions possibles
self.Q = np.zeros((self.c * self.l, 4))
#Probabilite d'exploration
self.exploration = 0.05
def __init__(self, params):
"""initialisation de l'agent"""
Agent.__init__(self, params)
self.l = int(params['ligne'])
self.c = int(params['colonne'])
#strategie initiale (au depart l'agent ne connait pas les deplacements optimaux a effectuer)
#tableau de taille nombre d'etat * nombre d'actions possibles
self.Q = np.zeros((self.l * self.c, 4))
#Probabilite d'exploration
self.exploration = 0.05
self.alpha = 0.6 #coefficient optimises
self.gamma = 1
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
#Nombre de lignes
self.l = int(params['ligne'])
#Nombre de colonnes
self.c = int(params['colonne'])
#strategie initiale (au depart l'agent ne connait pas les deplacements optimaux a effectuer)
#tableau de taille nombre d'etat * nombre d'actions possibles
self.Q = np.zeros((self.c * self.l, 4))
#Probabilite d'exploration
self.exploration = 0.05
#Coefficients pour la formule de Q
self.alpha = 1
self.gamma = 1
#vitesse de convergence initialisee (etape a partir de laquelle les recompenses totales sont superieures a 60
self.vitesse = 0
#numero de l'episode qui vient de finir
self.ep = 0
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
def __init__(self, params):
"""See documentation in the base class"""
Agent.__init__(self, params)
self.final_state = EnvironmentGrid2D(params).terminal_state
