TensorFlow多元线性回归占位符和矩阵乘法误差

import tensorflow as tf
import numpy as np

# 生成一些示例数据
np.random.seed(0)
X_data = np.random.rand(100, 3)  # 100个样本，每个样本有3个特征
y_data = X_data @ np.array([1.5, -2.0, 1.0]) + 0.5  # 线性关系 y = 1.5*x1 - 2.0*x2 + 1.0*x3 + 0.5

# 定义占位符
X = tf.placeholder(tf.float32, shape=[None, 3], name='X')
y = tf.placeholder(tf.float32, shape=[None], name='y')

# 定义权重和偏置
W = tf.Variable(tf.random_normal([3, 1]), name='weights')
b = tf.Variable(tf.random_normal([1]), name='bias')

# 预测值
y_pred = tf.add(tf.matmul(X, W), b)

# 损失函数（均方误差）
loss = tf.reduce_mean(tf.square(y_pred - y))

# 优化器
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01).minimize(loss)

# 初始化变量
init = tf.global_variables_initializer()

# 训练模型
with tf.Session() as sess:
    sess.run(init)
    
    for epoch in range(1000):
        _, l = sess.run([optimizer, loss], feed_dict={X: X_data, y: y_data})
        if (epoch + 1) % 100 == 0:
            print(f'Epoch {epoch + 1}, Loss: {l}')
    
    # 获取训练后的权重和偏置
    W_value, b_value = sess.run([W, b])
    print(f'Trained weights: {W_value.flatten()}')
    print(f'Trained bias: {b_value[0]}')

# 预测
with tf.Session() as sess:
    sess.run(init)
    y_pred_value = sess.run(y_pred, feed_dict={X: X_data})
    print(f'Predictions: {y_pred_value.flatten()}')