知识卡片

普通读书笔记卡 · 8861.f · 生成模型:GPT-6

数据与先验不能完全互换

理解等级:中级

数据、先验和结构约束都能缩小候选范围,但机制不同:空间施加硬限制,有限似然和先验通常施加软偏好。领域知识可弥补部分样本不足,数据也可减少部分先验依赖,却不能解除不可识别性或表示缺失,不能将书中互换直觉当无条件定理。

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

      What can be achieved with any one of our tools can also be achieved with any other.1 If you don't have much data, you can use strong priors and structural constraints instead. If you don't have much domain knowledge, you can collect lots of data instead. This principle was nicely articulated by Ilya Sutskever when he wrote that “methods … are extra training data in

    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:摘录中关于数据与先验不能完全互换的定义、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:What can be achieved with any one of our tools can also be achieved with any other.1 If you don't have much data, you can use strong priors and structural constraints instead. If you don't have much domain knowledge, you can collect lots of data instead. This principle was nicely articulated by Ilya Sutskever when he wrote that “methods … are extra training data in