def __init__( self, model=None, replay=None, exploration=None, actor_updater=None, critic_updater=None ): self.model = model or default_model() self.replay = replay or replays.Buffer() self.exploration = exploration or explorations.NormalActionNoise() self.actor_updater = actor_updater or \ updaters.DeterministicPolicyGradient() self.critic_updater = critic_updater or \ updaters.DeterministicQLearning()
def default_exploration(): return explorations.NormalActionNoise(scale=0.1, start_steps=10000)