Keras optimizer.get_updates(), K.function()のパラメータで[]は何ですか？ def critic_optimizer(self): #placeholder for target??? disc_reward = K.placeholder(shape=(None,1)) #output of critic value = self.critic.output #MSE error loss = K.mean(K.square(disc_reward - value)) optimizer = Adam(lr=self.critic_lr) #what is the second [] parameter??? updates = optimizer.get_updates(self.critic.trainable_weights, [], loss) #[] is an empty list for outputs? train = K.function([self.critic.input, disc_reward], [], updates=updates) return train A2CKeras[] def train_model(self, state, action, target, advantages): self.actor.optimizer([state, action, advantages]) self.critic.optimizer([state, target]) KerasTensorflow

<

Function

* * *

Tensorflow
<
<


`outputs``[]`
<


return fetched[:len(self.outputs)]


0




self.critic.optimizer([state, target])


****