矢量化的L1损失？
我正在阅读一篇论文，它提到他们正在使用 vectorized L1损失，这在编码方面意味着什么？与常规L1损失有区别吗？从纸上，这是公式：
查看lf，如果我要编码它，是否看起来像常规的L1损失？
import torch.nn.functional as F
loss_f = F.l1_loss(D_t, warped * D_t)
loss_f.backward()
如果是这样，为什么要特别提及 vectorized L1损失 ？我想念什么？
I was reading a paper and it mentioned that they were using vectorized L1 loss, what does that mean in terms of coding? Is there any difference from regular L1 loss? From the paper, this is the formula:
Looking at Lf, if I were to code it, wasn't it just look like the regular L1 loss?
import torch.nn.functional as F
loss_f = F.l1_loss(D_t, warped * D_t)
loss_f.backward()
if so, why the mention of vectorized L1 loss specifically? What did I miss?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
矢量化是计算机/数据科学中广泛使用的概念。在这里，它指的是计算L1损耗的方法，但是结果计算仍然相同。向量数学通常用作加快代码的方法，您可以阅读更多有关它的信息在这里。
Vectorization is a widely used concept in computer/data science. Here it refers to a method of computing the L1 loss, but the resulting calculation is still the same. Vector math is often used as a method to speed up code, and you can read a bit more about it here.