python - 回归损失函数不正确

Question

我正在尝试一个基本的平均示例，但验证和损失不匹配，并且如果我增加训练时间，网络将无法收敛。我正在训练一个具有 2 个隐藏层的网络，每个隐藏层宽度为 500 个单位，范围为 [0,9] 范围内的三个整数，学习率为 1e-1，Adam，批量大小为 1，丢失 3000 次迭代，并验证每个隐藏层100 次迭代。如果标签和假设之间的绝对差异小于阈值，这里我将阈值设置为1，我认为是正确的。有人可以告诉我这是损失函数选择的问题、Pytorch 的问题还是我正在做的事情吗？下面是一些图:

val_diff = 1
acc_diff = torch.FloatTensor([val_diff]).expand(self.batch_size)

验证期间循环 100 次:

num_correct += torch.sum(torch.abs(val_h - val_y) < acc_diff)

在每个验证阶段后附加:

validate.append(num_correct / total_val)

以下是(假设和标签)的一些示例:

[...(-0.7043088674545288, 6.0), (-0.15691305696964264, 2.6666667461395264),
 (0.2827358841896057, 3.3333332538604736)]

我尝试了 API 中通常用于回归的六个损失函数:

                                                    torch.nn.L1Loss(size_average=False)

                                                                      torch.nn.L1Loss()

                                                     torch.nn.MSELoss(size_average=False)

                                                                  torch.nn.MSELoss()

                                            torch.nn.SmoothL1Loss(size_average=False)

                                                               torch.nn.SmoothL1Loss()

谢谢。

网络代码:

class Feedforward(nn.Module):
    def __init__(self, topology):
        super(Feedforward, self).__init__()
        self.input_dim     = topology['features']
        self.num_hidden    = topology['hidden_layers']
        self.hidden_dim    = topology['hidden_dim']
        self.output_dim    = topology['output_dim']
        self.input_layer   = nn.Linear(self.input_dim, self.hidden_dim)
        self.hidden_layer  = nn.Linear(self.hidden_dim, self.hidden_dim)
        self.output_layer  = nn.Linear(self.hidden_dim, self.output_dim)
        self.dropout_layer = nn.Dropout(p=0.2)


    def forward(self, x):
        batch_size = x.size()[0]
        feat_size  = x.size()[1]
        input_size = batch_size * feat_size

        self.input_layer = nn.Linear(input_size, self.hidden_dim).cuda()
        hidden = self.input_layer(x.view(1, input_size)).clamp(min=0)

        for _ in range(self.num_hidden):
            hidden = self.dropout_layer(F.relu(self.hidden_layer(hidden)))

        output_size = batch_size * self.output_dim
        self.output_layer = nn.Linear(self.hidden_dim, output_size).cuda()
        return self.output_layer(hidden).view(output_size)

训练代码:

def train(self):
    if self.cuda:
        self.network.cuda()

    dh        = DataHandler(self.data)
    # loss_fn = nn.L1Loss(size_average=False)
    # loss_fn = nn.L1Loss()
    # loss_fn = nn.SmoothL1Loss(size_average=False)
    # loss_fn = nn.SmoothL1Loss()
    # loss_fn = nn.MSELoss(size_average=False)
    loss_fn   = torch.nn.MSELoss()
    losses    = []
    validate  = []
    hypos     = []
    labels    = []
    val_size  = 100
    val_diff  = 1
    total_val = float(val_size * self.batch_size)

    for i in range(self.iterations):
        x, y = dh.get_batch(self.batch_size)
        x = self.tensor_to_Variable(x)
        y = self.tensor_to_Variable(y)

        self.optimizer.zero_grad()
        loss = loss_fn(self.network(x), y)
        loss.backward()
        self.optimizer.step()

Answer 1

看来您误解了 pytorch 中图层的工作原理，这里有一些提示:

• 当您这样做时，在您的转发中nn.Linear(...)您正在定义新层，而不是使用您在网络中预定义的层__init__ 。因此，它无法学习任何东西，因为权重不断重新初始化。

• 您不需要调用.cuda()里面net.forward(...)因为您已经在 train 中复制了 GPU 上的网络通过调用 self.network.cuda()

• 理想情况下是net.forward(...)输入应该直接具有第一层的形状，因此您不必修改它。在这里你应该有x.size() <=> Linear -- > (Batch_size, Features) .

您的转发应该与此接近:

def forward(self, x):
    x = F.relu(self.input_layer(x))
    x = F.dropout(F.relu(self.hidden_layer(x)),training=self.training)
    x = self.output_layer(x)
    return x