机器学习总结

1.BN【batch normalization】

https://zhuanlan.zhihu.com/p/93643523

减少 

2.L1L2正则化

l1:稀疏

l2:权重减小

3.泛化误差

训练误差计算了训练集的误差，而泛化误差是计算全集的误差。

4.dropout

训练过程中神经元p的概率失活

一文彻底搞懂深度学习：正则化（Regularization）-CSDN博客

5.train里该有什么

# 设置损失函数和优化器

    criterion = torch.nn.MSELoss()

    optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(

        optimizer, mode='min', factor=0.5, patience=patience//4, verbose=True

    )

for epoch in range(epochs):

        # 训练阶段

        model.train()

        epoch_loss = 0.0

        batch_count = 0

       

        for inputs, targets in train_loader:

              optimizer.zero_grad()

            outputs = model(inputs)

            loss = criterion(outputs, targets)

            loss.backward()

            optimizer.step()

//定期验证