激活函数
- relu:tf.nn.relu()
- sigmoid:tf.nn,sigmoid()
- tanh:tf.nn.tnah()
NN复杂度
层数 = 隐藏层层数 + 一个输出层
总参数 = 总权重参数 + 总偏置参数
损失函数
预测值(y)与已知答案(y_)的差距(loss最小)
衡量标准:MSE、CE…、自定义损失函数
模型的保存:
save model:
saver = tf.train.Saver();
with tf.Session() as sess:
for i in range(STEPS):
if i % 轮数 == 0:
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step = global_step);
加载model:
with tf.Session() as sess:
ckpt = td.train.get_checkpoint_path:
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path);