tensorflow均方误差损失函数预测酸奶价格_计算两个tensor的均方误差-优快云博客

本文介绍使用TensorFlow的均方误差(MSE)损失函数进行酸奶价格预测的详细过程，包括数据准备、模型训练及参数调整。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

tensorflow均方误差损失函数

引用API：tensorflow.kreas.losses.MSE

$MSE(y,y')=\tfrac{\sum _{i=1}^{n}(y_{i}-y_{i}')^{2}}{n}$

均方误差（Mean Square Erroe）是回归问题最常用的损失函数，回归问题解决的是对具体数值的预测，而不是预测某个事先定义好的分类

则，lose_mean = tf.reduce_mean(tf.square(y_, y)) #y_为真实值，y_为计算出的预测值

tensorflow均方误差损失函数预测酸奶价格

import tensorflow as tf
import numpy as np

seed = 23456 #定义种子数值

rdm = np.random.RandomState(seed) #生成[0, 1)之间的小数,由于固定seed数值，每次生成的数值都一样

x = rdm.rand(32, 2) #生成32行2列的输入特征x

np.set_printoptions(threshold=6) #概略显示

print("32行2列的特征xs为：", x)  # 打印输入特征，概略显示5行

y_ = [[x1 + x2 + (rdm.rand()  / 10.0 - 0.05)] for (x1, x2) in x] #从x中取出x1,x2相加，并且加上-0.05到+0.05随机噪声，得到标准答案y_

x = tf.cast(x, dtype = tf.float32) #转换x数据类型

print("转换数据类型后x:", x) #打印转换书籍类型后的x

w1 = tf.Variable(tf.random.normal([2, 1], stddev = 1, seed = 1)) #初始化参数w1，生成两行一列的可训练参数

print("w1为：", w1)

epoch = 15000 #迭代次数为15000次

lr = 0.002 #学习率为0.002

for epoch in range(epoch):
    with tf.GradientTape() as tape: #用with上下文管理器调用GradientTape()方法
        y = tf.matmul(x, w1) #求前向传播计算结果y
        loss_mse = tf.reduce_mean(tf.square(y_ - y)) #求均方误差loss_mean
# print("y为：", y)
# print("loss_mean为：", loss_mse)
    grads = tape.gradient(loss_mse, w1) #损失函数对待训练参数w1求偏导
    w1.assign_sub(lr * grads) #更新参数w1

    if epoch % 500 == 0: #每迭代500轮
        print("After %d training steps, w1 is " % epoch) #每迭代500轮打印当前w1参数
        print(w1.numpy(), "\n") #打印当前w1参数

print("Final w1 is :", w1.numpy()) #打印最终w1

结果为：

E:\Anaconda3\envs\TF2\python.exe C:/Users/Administrator/PycharmProjects/untitled8/酸奶价格预测.py
32行2列的特征xs为： [[0.32180029 0.32730047]
[0.92742231 0.31169778]
[0.16195411 0.36407808]
...
[0.83554249 0.59131388]
[0.2829476 0.05663651]
[0.2916721 0.33175172]]
转换数据类型后x: tf.Tensor(
[[0.3218003 0.32730046]
[0.9274223 0.31169778]
[0.1619541 0.36407807]
...
[0.8355425 0.5913139 ]
[0.2829476 0.05663651]
[0.29167208 0.33175173]], shape=(32, 2), dtype=float32)
w1为： <tf.Variable 'Variable:0' shape=(2, 1) dtype=float32, numpy=
array([[-0.8113182],
[ 1.4845988]], dtype=float32)>
After 0 training steps, w1 is
[[-0.8093924]
[ 1.4857364]]

After 500 training steps, w1 is
[[-0.17545831]
[ 1.7493846 ]]

After 1000 training steps, w1 is
[[0.12224181]
[1.7290779 ]]

After 1500 training steps, w1 is
[[0.30012047]
[1.639089 ]]

After 2000 training steps, w1 is
[[0.42697424]
[1.5418574 ]]

After 2500 training steps, w1 is
[[0.5261725]
[1.4540646]]

After 3000 training steps, w1 is
[[0.6068525]
[1.378847 ]]

After 3500 training steps, w1 is
[[0.6734775]
[1.3155416]]

After 4000 training steps, w1 is
[[0.72881055]
[1.2626004 ]]

After 4500 training steps, w1 is
[[0.77486306]
[1.2184267 ]]

After 5000 training steps, w1 is
[[0.81322 ]
[1.1816002]]

After 5500 training steps, w1 is
[[0.8451773]
[1.1509084]]

After 6000 training steps, w1 is
[[0.87180513]
[1.1253314 ]]

After 6500 training steps, w1 is
[[0.8939932]
[1.1040187]]

After 7000 training steps, w1 is
[[0.9124815]
[1.0862588]]

After 7500 training steps, w1 is
[[0.9278873]
[1.07146 ]]

After 8000 training steps, w1 is
[[0.94072455]
[1.0591285 ]]

After 8500 training steps, w1 is
[[0.9514216]
[1.0488532]]

After 9000 training steps, w1 is
[[0.96033514]
[1.040291 ]]

After 9500 training steps, w1 is
[[0.9677624]
[1.0331562]]

After 10000 training steps, w1 is
[[0.9739516]
[1.0272108]]

After 10500 training steps, w1 is
[[0.97910905]
[1.0222566 ]]

After 11000 training steps, w1 is
[[0.98340666]
[1.0181286 ]]

After 11500 training steps, w1 is
[[0.98698765]
[1.0146885 ]]

After 12000 training steps, w1 is
[[0.98997146]
[1.0118226 ]]

After 12500 training steps, w1 is
[[0.9924576]
[1.0094334]]

After 13000 training steps, w1 is
[[0.99452955]
[1.0074434 ]]

After 13500 training steps, w1 is
[[0.9962558]
[1.0057855]]

After 14000 training steps, w1 is
[[0.997694 ]
[1.0044047]]

After 14500 training steps, w1 is
[[0.9988928]
[1.0032523]]

Final w1 is : [[0.99988985]
[1.0022951 ]]

Process finished with exit code 0