为什么深度学习模型的预测结果为数据集均值--pytorch loss里input和target的维度必须相同

人间真正走过~

已于 2023-11-14 18:43:01 修改

阅读量636

点赞数 2

文章标签：深度学习均值算法人工智能

于 2023-11-07 10:38:46 首次发布

本文链接：https://blog.csdn.net/weixin_37198422/article/details/134261764

版权

如图，上边的为真值，下边的为预测值，可见预测值都是在30左右

一直debug不出来。后来发现是loss处写错了

用的torch.nn.MSELoss(pred,target)，其中pred的size是(batchsize,1), target的size是(batchsize)，此时会有警告但不报错：

media/cfs/user/.pylib/lib/python3.6/site-packages/torch/nn/modules/loss.py:445: UserWarning: Using a target size (torch.Size([2560])) that is different to the input size (torch.Size([2560, 1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
  return F.mse_loss(input, target, reduction=self.reduction)
/media/cfs/user/.pylib/lib/python3.6/site-packages/torch/nn/modules/loss.py:445: UserWarning: Using a target size (torch.Size([360])) that is different to the input size (torch.Size([360, 1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
  return F.mse_loss(input, target, reduction=self.reduction)

因此需要把pred的1那维给去掉，用squeeze或者reshape都行

我们做个简单的实验。下图中的3.1375才是正确的结果（因为MSELoss是对所有样本的mse求均值）。而对target进行squeeze后，input和target的size不再相同，所得结果2.450也是错的