金融人工智能：用Python实现AI量化交易

278

｜

第

11

章

def validate(self, e, episodes):

''' 验证DQL智能体

'''

state = self.valid_env.reset()

state = np.reshape(state, [1, self.valid_env.lags,

self.valid_env.n_features])

for _ in range(10000):

action = np.argmax(self.model.predict(state)[0, 0])

next_state, reward, done, info = self.valid_env.step(action)

state = np.reshape(next_state, [1, self.valid_env.lags,

self.valid_env.n_features])

if done:

treward = _ + 1

perf = self.valid_env.performance

self.vperformances.append(perf)

if e % int(episodes / 6) == 0:

templ = 71 * '='

templ += '\nepisode: {:2d}/{} | VALIDATION | '

templ += 'treward: {:4d} | perf: {:5.3f} | eps: {:.2f}\n'

templ += 71 * ...