← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8861.c · 生成模型:GPT-6

数据为候选增加兼容约束

理解等级:中级

在两参数基函数例中,一个非退化观测可约束参数平面的一条线;多个含噪观测的约束未必交于一点,需要按损失折中。更多有信息且代表性的数据可缩小不确定范围,但重复、偏置或不可识别的数据未必定位真解,不能把数量当充分条件。

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

      Each single datapoint adds one constraint to this problem, and each constraint shows up as a line in the heatmap for J [the line of solutions that satisfy the constraint y(i) = θ0x(i) + θ1 sin(x(i)) for that datapoint i]. The intersection of all these constraints pinpoints the setting of parameters that fits all that data. In this example, with five datapoints, there is no

    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:摘录中关于数据为候选增加兼容约束的定义、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:Each single datapoint adds one constraint to this problem, and each constraint shows up as a line in the heatmap for J [the line of solutions that satisfy the constraint y(i) = θ0x(i) + θ1 sin(x(i)) for that datapoint i]. The intersection of all these constraints pinpoints the setting of parameters that fits all that data. In this example, with five datapoints, there is no