def start_nn():
print("--------- 1.load data ------------")
feature, label, n_output = load_data("data.txt")
print("--------- 2.training ------------")
center, delta, w = bp_train(feature, label, 20, 5000, 0.008, n_output)
print("--------- 3.get prediction ------------")
result = get_predict(feature, center, delta, w)
# print("result：", (1 - err_rate(label, result)))
print("--------- 4.save model and result ------------")
save_model_result(center, delta, w, result)
return jsonify({"res": "success"})
alpha = min_lr
steps = 0
center = mat(np.random.rand(n_hidden, n))
delta = mat(np.random.rand(1, n_hidden))
w = mat(np.random.rand(n_hidden, n_output))
while steps <= 3000:
env.render()
pos, vel = discretization(env, obs)
state = normalize(np.array([pos, vel]).reshape(1, 2),
axis=1,
norm='max')
state = np.matrix(state)
a = get_action(state)
obs, reward, terminate, _ = env.step(a)
total_reward += abs(obs[0] + 0.5)
pos_, vel_ = discretization(env, obs)
q_table[pos][vel][a] = (1 - alpha) * q_table[pos][vel][a] + alpha * (
reward + gamma * np.max(q_table[pos_][vel_]))
center, delta, w, loss = bp_train(state, np.matrix(q_table[pos][vel]),
n_hidden, 1, 0.005, 3, center, delta,
w)
steps += 1
if terminate:
print("Finished after: " + str(episode) + " steps" + str(steps))
print("Cumulated Reward: " + str(total_reward))
print("Complete!")
break
while True:
env.render()
steps = 0
center = mat(np.random.rand(n_hidden, n))
delta = mat(np.random.rand(1, n_hidden))
w = mat(np.random.rand(n_hidden, n_output))
while steps <= 3000:
env.render()
#pos, vel = obs[0],obs[1]
state1 = normalize_state(obs)
state1 = np.matrix(state1)
a = get_action(state1)
#print('action',a)
obs, reward, terminate, _ = env.step(a)
total_reward += abs(obs[0] + 0.5)
state2 = normalize_state(obs)
state2 = np.matrix(state2)
predict = get_predict(state2, center, delta, w)
target[0][a] = (1 - alpha) * target[0][a] + alpha * (
reward + gamma * np.max(predict))
center, delta, w, loss = bp_train(state1, np.matrix(target), n_hidden,
1, 0.01, 3, center, delta, w)
steps += 1
if terminate:
print("Finished after: " + str(episode) + " steps" + str(steps))
print("Cumulated Reward: " + str(total_reward))
print("Complete!")
break
#while True:
#env.render()