$$MSE-均方误差$$

补充：reduce操作是用来减维度

---

一 TF2.0

---

1.1 类实现

class MeanSquaredError(tf.keras.losses.Loss):def call(self, y_true, y_pred):return tf.reduce_mean(tf.square(y_pred - y_true))

a = tf.constant([0., 0., 1., 1.])
b = tf.constant([1., 1., 1., 0.])

mse = MeanSquaredError()

mse(a,b)

1.2 函数式实现

def MeanSquaredError( y_true, y_pred):return tf.reduce_mean(tf.square(y_pred - y_true))

mse = MeanSquaredError(a,b)
mse

1.3 tf2.0的keras高阶API

mse = tf.keras.losses.MeanSquaredError()
loss = mse([0., 0., 1., 1.], [1., 1., 1., 0.])
print('Loss: ', loss.numpy())  # Loss: 0.75

---

二 Numpy实现