def test_nn():
req_data = request.get_json()
print("--------- 1.load data ------------")
dataTest = load_data(
[
req_data['x1']
, req_data['x2']
, req_data['x3']
, req_data['x4']
, req_data['x5']
, req_data['x6']
, req_data['x7']
, req_data['x8']
, req_data['x9']
, req_data['x10']
, req_data['x11']
, req_data['x12']
, req_data['x13']
, req_data['x14']
, req_data['x15']
, req_data['x16']
, req_data['x17']
, req_data['x18']
, req_data['x19']
]
)
print("--------- 2.load model ------------")
center, delta, w = load_model("messidor_center.txt", "messidor_delta.txt", "messidor_weight.txt")
print("--------- 3.get prediction ------------")
result = get_predict(dataTest, center, delta, w)
print('result', result)
print("--------- 4.save result ------------")
res = save_predict(result)
return jsonify({ "res": res })
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"})
def get_action(state):
if np.random.uniform(low=0, high=1) < epsilon:
return np.random.choice(env.action_space.n)
predict = get_predict(state, center, delta, w)
return np.argmax(predict)
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()
delta = get_model(file_delta)
w = get_model(file_w)
return center, delta, w
def save_predict(pre):
m = shape(pre)[0]
result = []
for i in range(m):
if (pre[i, 0] < 0.5):
pre[i, 0] = 0
else:
pre[i, 0] = 1
result.append(str(pre[i, 0]))
print(result)
return result
if __name__ == "__main__":
print("--------- 1.load data ------------")
dataTest = load_data(x, y)
print(dataTest)
print("--------- 2.load model ------------")
center, delta, w = load_model("messidor_center.txt", "messidor_delta.txt",
"messidor_weight.txt")
print("--------- 3.get prediction ------------")
result = get_predict(dataTest, center, delta, w)
print("--------- 4.save result ------------")
save_predict(result)
