tf.keras.losses.MeanSquaredError 均方差 示例

import tensorflow as tf
import numpy as np

均方差 常用于回归

计算标签和预测之间的误差平方均值

y_true = [[0., 1.], [0., 0.]]
y_pred = [[1., 1.], [1., 0.]]
mse = tf.keras.losses.MeanSquaredError()  # 均方差损失函数
mse(y_true, y_pred).numpy()

0.5

y_true = [[0., 1.], [0., 0.], [9., 15.]]
y_pred = [[1., 1.], [1., 0.], [1., 3.5]]
mse = tf.keras.losses.MeanSquaredError()
mse(y_true, y_pred).numpy()

33.041668