构建网络

x = tf.placeholder(tf.float32, shape=[None, 4])
y = tf.placeholder(tf.float32, shape=[None, 3])
l1 = tf.layers.dense(x, 128, tf.nn.tanh, name=”l1”)
l2 = tf.layers.dense(l1, 128, tf.nn.tanh, name=”l2”)
l3 = tf.layers.dense(l2, 128, tf.nn.tanh, name=”l3”)
l4 = tf.layers.dense(l3, 256, tf.nn.tanh, name=”l4”)
l5 = tf.layers.dense(l4, 256, tf.nn.tanh, name=”l5”)
l6 = tf.layers.dense(l5, 256, tf.nn.tanh, name=”l6”)
l7 = tf.layers.dense(l6, 256, tf.nn.tanh, name=”l7”)
y_pred = tf.layers.dense(l7, 3)

定义损失

loss = tf.reduce_mean(tf.square(y - y_pred))
optimizer = tf.train.AdamOptimizer(1e-4).minimize(loss)

初始化变量

init = tf.global_variables_initializer()

以下为训练神经网络的部分代码，训练时喂入的数据分别为x和y。

sess.run(init)
train_feed_dict = {
x: dataset_X,
y: dataset_Y
}
saver.save(sess, model_path)
print(“训练结束，模型保存到{}”.format(model_path))

以下测试代码，测试时喂入的数据也分别为x和y。

saver.restore(sess, model_path1)
print(“从{}载入模型”.format(model_path1))
test_feed_dict = {
x: dataset_X,
y: dataset_Y
}
np.set_printoptions(suppress=True)
prediction = sess.run(y_pred, feed_dict=test_feed_dict)

目标：调用训练好的模型，输入x，输出y