next_state = np.reshape(next_state, [1, state_size])
# Remember the previous state, action, reward, and done
agent.remember(state, action, reward, next_state, done)
# make next_state the new current state for the next frame.
state = next_state
# done becomes True when the game ends
# ex) The agent drops the pole
if done:
agent.update_target_model()
print("episode: {}/{}, score: {}, e: {:.2}"
.format(e, EPISODES, time, agent.epsilon))
break
if len(agent.memory) > batch_size:
# train the agent with the experience of the episode
# loss = agent.replay(batch_size)
agent.replay(batch_size)
# Logging training loss every 10 timesteps
# if time % 10 == 0:
# print("episode: {}/{}, time: {}, loss: {:.4f}"
# .format(e, EPISODES, time, loss))
# if e % 10 == 0:
# agent.save("cartpole-dqn.h5")
# # --------------------------------------------------------------------------------------------------------------------------------------------------------
# SARSA Q-TABLE UPDATE EQUATIION
# Q(s,a) = Q(s,a) + alpha(R + gamma*Q(s`,a`) - Q(s,a)) => regardless of negetive and positive value we'll use mse as our loss function
# KERAS-RL
ENV_NAME = 'CartPole-v0'
