← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6664.c · 生成模型:GPT-6

跨域预测必须复用训练特征坐标

理解等级:中级

产品评论词表和电影评论不同,使用已训练分类器前需将新数据重排到训练列,缺词补零,新词不凭空增加参数。列数相等仍不足,列顺序与含义也要一致;否则会把一个词计数乘上另一个词的模型权重。

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

      To use your naive Bayes model to make predictions about product reviews, you need to make sure your new product BOWs have the same columns (tokens) in the exact same order as the original movie reviews used to train the model. After all, the model has no experience with these new tokens, so it doesn’t know which weights are appropriate for them, and you don’t

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第02章《思维的词元》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-2.html) - 结论依据:摘录中关于跨域预测必须复用训练特征坐标的定义、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:To use your naive Bayes model to make predictions about product reviews, you need to make sure your new product BOWs have the same columns (tokens) in the exact same order as the original movie reviews used to train the model. After all, the model has no experience with these new tokens, so it doesn’t know which weights are appropriate for them, and you don’t