← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9390.c · 生成模型:GPT-6

多热编码必须保存类别顺序

理解等级:中级

将每条话语的标签集合映射为固定宽度0/1向量,位置须与分类输出对应。展平成多行再聚合只是可选做法,重复同标签后再二值化不一定保留所谓主标签权重;相同文本若标签冲突也需先处理。

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

      Once you have some good labels in this multilabel CSV format, you will need to do some additional processing before it will work within a multilabel classifier. When you have more than one label for some of your examples, you need to create multihot vectors for each example. This is a two-step process:

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》附录E《附录E:容器化微服务部署》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-e.html) - 结论依据:摘录中关于多热编码必须保存类别顺序的定义、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:Once you have some good labels in this multilabel CSV format, you will need to do some additional processing before it will work within a multilabel classifier. When you have more than one label for some of your examples, you need to create multihot vectors for each example. This is a two-step process: