← 返回刚才阅读的位置

知识卡片

结构图卡 · 8861.g · 生成模型:GPT-6

约束真解搜索的三种来源

理解等级:中级

模型空间、观测证据与先验各自回答能表达什么、与数据是否兼容、事先偏好什么。优化从共同约束中选择候选,泛化还要用独立未见数据检验。

结构图:

flowchart TD
 A[假设空间的硬边界] --> D[可搜索候选]
 B[训练数据的拟合损失] --> E[候选评分]
 C[先验或正则项] --> E
 D --> F[优化选择函数]
 E --> F
 F --> G[独立验证与测试]
 G --> H[目标分布下的泛化证据]
这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      The hypothesis space constrains which solutions we can possibly find. If the true solution is not in our hypothesis space, then no amount of data or priors can help us find it. This situation is like the old joke of a drunk man looking for his lost keys under a lamppost [137]. “Why are you looking there,” a cop asks. “Because this is where the

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第11章《泛化问题》(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/c011.xhtml) - 结论依据:摘录中关于约束真解搜索的三种来源的定义、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 hypothesis space constrains which solutions we can possibly find. If the true solution is not in our hypothesis space, then no amount of data or priors can help us find it. This situation is like the old joke of a drunk man looking for his lost keys under a lamppost [137]. “Why are you looking there,” a cop asks. “Because this is where the