MaxA = a
return MaxV,MaxA
def accionRandom(a,grid = Grid.grid_estandar(), eps=EPS):
p = np.random.random()
if p < (1.0 - eps):
return a
else:
return np.random.choice(grid.posiblesAccionesBasicas())
if __name__ == "__main__":
grid = Grid.grid_negativo(-0.1)
todos_estados = grid.todos_estados()
Q = {}
alpha_personal = {}
for s in todos_estados:
Q[s] = {}
alpha_personal[s] = {}
for a in grid.posiblesAccionesBasicas():
Q[s][a] = 0
alpha_personal[s][a] = 1.0
t = 1.0
def MaxQ(Q_s):
MaxV = float('-inf')
MaxA = None
for k, v in Q_s.items():
if v > MaxV:
MaxV = v
MaxA = k
return (MaxV, MaxA)
if __name__ == "__main__":
grid = Grid.grid_negativo(penalizacion=-0.2)
Q, politica, historialRetornos = iniciarValores(grid)
print("Recompensas")
mostrar_valores(grid.recompensas, grid)
print("\nPolítica Inicial")
mostrar_politica(politica, grid)
print("\nValores Iniciales")
V = {}
for s in politica:
V[s] = MaxQ(Q[s])[0]
mostrar_valores(V, grid)
episodios = 10000