知识卡片

普通读书笔记卡 · 6486.b · 生成模型:GPT-6

正则强度与核尺度有别

理解等级:高级

SVM的C控制分类误差惩罚相对正则项的权重;某些核的gamma控制相似性随距离衰减的尺度。两者作用不同,C大不保证在每个数据集上得到更小的实际几何间隔,参数需在固定验证协议下共同比较。

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 value of parameter C is a regularization parameter. A higher value of C represents a smaller margin, while a lower value of C generates a more significant margin. The effectiveness of the regularization parameter is inversely proportional to the value of C. Size of the cache is defined in terms of MB based on the type of kernels applied.

    参考来源

    - 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第02章《文本挖掘与垃圾邮件过滤》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications/_epub-src/OEBPS/Text/9789815136746-C2.html) - 结论依据:摘录中关于正则强度与核尺度有别的定义、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 value of parameter C is a regularization parameter. A higher value of C represents a smaller margin, while a lower value of C generates a more significant margin. The effectiveness of the regularization parameter is inversely proportional to the value of C. Size of the cache is defined in terms of MB based on the type of kernels applied.