当代人工智能 实验一 使用神经网络解决三好学生问题

三好学生代码（TensorFlow 1.x）：
 

import tensorflow as tf

x1 = tf.placeholder(dtype=tf.float32)
x2 = tf.placeholder(dtype=tf.float32)
x3 = tf.placeholder(dtype=tf.float32)

w1 = tf.Variable(0.1, dtype=tf.float32)
w2 = tf.Variable(0.1, dtype=tf.float32)
w3 = tf.Variable(0.1, dtype=tf.float32)

n1 = x1 * w1
n2 = x2 * w2
n3 = x3 * w3

y = n1 + n2 + n3

sess = tf.Session()

init = tf.global_variables_initializer()
sess.run(init)

result = sess.run([x1, x2, x3, w1, w2, w3, y], feed_dict={x1: 90, x2: 80, x3: 70})
print(result)

yTrain = tf.placeholder(dtype=tf.float32)

loss = tf.abs(y - yTrain)
optimizer = tf.train.RMSPropOptimizer(0.001)

train = optimizer.minimize(loss)

result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 90, x2: 80, x3: 70, yTrain: 85})
print(result)

result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 98, x2: 95, x3: 87, yTrain: 96})
print(result)

for i in range(5000):
	result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 90, x2: 80, x3: 70, yTrain: 85})
	print(result)

result = sess.run([train, x1, x2, x3, w1, w2, w3, y, yTrain, loss], feed_dict={x1: 98, x2: 95, x3: 87, yTrain: 96})
print(result)

若是TensorFlow 2.x的版本，需要修改代码，运行如下代码：

import tensorflow as tf
# 定义变量
w1 = tf.Variable(0.1, dtype=tf.float32)
w2 = tf.Variable(0.1, dtype=tf.float32)
w3 = tf.Variable(0.1, dtype=tf.float32)
# 定义模型函数
def model(x1, x2, x3):
    n1 = x1 * w1
    n2 = x2 * w2
    n3 = x3 * w3
    y = n1 + n2 + n3
    return y
# 定义损失函数
def loss_function(y_pred, y_true):
    return tf.abs(y_pred - y_true)
# 选择优化器
optimizer = tf.optimizers.RMSprop(0.001)
# 训练步骤函数
def train_step(x1, x2, x3, y_true):
    with tf.GradientTape() as tape:
        y_pred = model(x1, x2, x3)
        loss = loss_function(y_pred, y_true)
    gradients = tape.gradient(loss, [w1, w2, w3])
    optimizer.apply_gradients(zip(gradients, [w1, w2, w3]))
    return loss
# 示例数据和训练
x1_data = tf.constant([90, 98], dtype=tf.float32)
x2_data = tf.constant([80, 95], dtype=tf.float32)
x3_data = tf.constant([70, 87], dtype=tf.float32)
y_train_data = tf.constant([85, 96], dtype=tf.float32)
# 运行训练循环
for i in range(5000):
    loss = train_step(x1_data, x2_data, x3_data, y_train_data)
    if i % 100 == 0:
        print(f"Iteration {i}, Loss: {loss.numpy()}")
# 检查最终的权重和损失
y_pred = model(x1_data, x2_data, x3_data)
final_loss = loss_function(y_pred, y_train_data)
print(f"Final weights: {w1.numpy()}, {w2.numpy()}, {w3.numpy()}")
print(f"Final loss: {final_loss.numpy()}")

分析权重代码（TensorFlow版本为2.x）：

import tensorflow as tf
# 定义变量
w1 = tf.Variable(0.1, dtype=tf.float32)
w2 = tf.Variable(0.1, dtype=tf.float32)
w3 = tf.Variable(0.1, dtype=tf.float32)
w4 = tf.Variable(0.1, dtype=tf.float32)
# 定义模型函数
def model(x1, x2, x3,x4):
    n1 = x1 * w1
    n2 = x2 * w2
    n3 = x3 * w3
    n4 = x4 * w4
    y = n1 + n2 + n3 + n4
    return y
# 定义损失函数
def loss_function(y_pred, y_true):
    return tf.abs(y_pred - y_true)
# 选择优化器
optimizer = tf.optimizers.RMSprop(0.001)
# 训练步骤函数
def train_step(x1, x2, x3, x4,y_true):
    with tf.GradientTape() as tape:
        y_pred = model(x1, x2, x3,x4)
        loss = loss_function(y_pred, y_true)
    gradients = tape.gradient(loss, [w1, w2, w3,w4])
    optimizer.apply_gradients(zip(gradients, [w1, w2, w3,w4]))
    return loss
# 示例数据和训练
x1_data = tf.constant([90, 80,90], dtype=tf.float32)
x2_data = tf.constant([80, 90,100], dtype=tf.float32)
x3_data = tf.constant([80, 100,70], dtype=tf.float32)
x4_data = tf.constant([70, 90,100], dtype=tf.float32)
y_train_data = tf.constant([84, 86,92], dtype=tf.float32)
# 运行训练循环
for i in range(5000):
    loss = train_step(x1_data, x2_data, x3_data,x4_data, y_train_data)
    if i % 100 == 0:
        print(f"Iteration {i}, Loss: {loss.numpy()}")
# 检查最终的权重和损失
y_pred = model(x1_data, x2_data, x3_data,x4_data)
final_loss = loss_function(y_pred, y_train_data)
print(f"Final weights: {w1.numpy()}, {w2.numpy()}, {w3.numpy()},{w4.numpy()}")
print(f"Final loss: {final_loss.numpy()}")

最终的权重为：

ω1=0.5046337842941284

ω2= 0.25738802552223206

ω3=0.10077429562807083

ω4=0.13746078312397003

作者：henu  21级 计算机与信息工程学院 空午