知识卡片
重建损失决定保真偏好
平方重建误差是Σi(x̂i−xi)²,直接累加逐像素偏差,不显式比较边缘或纹理的联合统计。容量有限时,损失规定优先保留哪些变化。可加入感知结构损失,但它也会改变“重建好”的含义,需按目标检验而非只看像素误差。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 表示学习
原书出处 · 1 处
- · 查看原书依据
The idea is to find a lower dimensional representation of the data from which we are able to reconstruct the data. An autoencoder is fine with throwing away any redundant features of the raw data but is not happy with actually losing information. The particular loss function determines what kind of information is preferred when the bottleneck cannot support preserving all the information and a
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第30章《表示学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c030b.xhtml)
- 结论依据:摘录中关于重建损失决定保真偏好的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:The idea is to find a lower dimensional representation of the data from which we are able to reconstruct the data. An autoencoder is fine with throwing away any redundant features of the raw data but is not happy with actually losing information. The particular loss function determines what kind of information is preferred when the bottleneck cannot support preserving all the information and a