Tensorflow中的加权训练示例

import tensorflow as tf

# 创建数据集
data = [
    (tf.constant([1.0, 2.0]), tf.constant(0.0)),
    (tf.constant([3.0, 4.0]), tf.constant(1.0)),
    (tf.constant([5.0, 6.0]), tf.constant(0.0)),
    (tf.constant([7.0, 8.0]), tf.constant(1.0))
]

dataset = tf.data.Dataset.from_generator(lambda: data, (tf.float32, tf.float32))

# 定义模型
model = tf.keras.Sequential([
    tf.keras.layers.Dense(1, activation='sigmoid')
])

# 定义损失函数和优化器
loss_fn = tf.keras.losses.BinaryCrossentropy()
optimizer = tf.keras.optimizers.SGD()

# 定义加权训练
def weighted_loss(y_true, y_pred, weights):
    loss = loss_fn(y_true, y_pred)
    weighted_loss = tf.reduce_mean(loss * weights)
    return weighted_loss

# 定义训练步骤
@tf.function
def train_step(inputs, labels, weights):
    with tf.GradientTape() as tape:
        predictions = model(inputs)
        loss = weighted_loss(labels, predictions, weights)
    gradients = tape.gradient(loss, model.trainable_variables)
    optimizer.apply_gradients(zip(gradients, model.trainable_variables))

# 进行加权训练
weights = tf.constant([1.0, 2.0, 1.0, 2.0])  # 样本的权重
epochs = 10

for epoch in range(epochs):
    for inputs, labels in dataset:
        train_step(inputs, labels, weights)

# 使用训练好的模型进行预测
inputs = tf.constant([[1.0, 2.0], [3.0, 4.0]])
predictions = model(inputs)
print(predictions)