知识卡片
缺失值任务涵盖多种轴
补洞、灰度上色和未来帧预测都可理解为沿空间、通道或时间轴补充缺失信息。哪些数据可见、缺失规则和预测损失决定任务难度。目标常有多种合理答案,单一像素回归容易偏向条件平均,不能将补全结果当作唯一真实恢复。
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
Imputation is a special case of self-supervised learning, where the prediction targets are missing elements of the input data. For example, predicting missing pixels is an imputation problem, as is colorizing a black and white photo (i.e., predicting missing color channels). Figure 30.7 gives several examples of these imputation tasks.
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:摘录中关于缺失值任务涵盖多种轴的定义、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:Imputation is a special case of self-supervised learning, where the prediction targets are missing elements of the input data. For example, predicting missing pixels is an imputation problem, as is colorizing a black and white photo (i.e., predicting missing color channels). Figure 30.7 gives several examples of these imputation tasks.