知识卡片

普通读书笔记卡 · 6664.b · 生成模型:GPT-6

训练集成绩不能代表泛化效果

理解等级:中级

书中用电影评论训练后又在同一矩阵上得到约93%的分类一致率,是训练表现。需另留独立样本并给多数类基准、类别分布与混淆情况,才能判断任务质量;不能以训练集准确率直接比较另一个未同协议评估的规则系统。

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

      >>> (movies.pred_ispos == ... movies.senti_ispos).sum() / len(movies) 0.9344648750589345 #1 #1 You got the “thumbs up” rating correct 93% of the time. This is a pretty good start at building a sentiment analyzer, with only a few lines of code (and a lot of data). You did not have to guess at the sentiment associated with a list of 7,500 words and hardcode them into an

    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:摘录中关于训练集成绩不能代表泛化效果的定义、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:>>> (movies.pred_ispos == ... movies.senti_ispos).sum() / len(movies) 0.9344648750589345 #1 #1 You got the “thumbs up” rating correct 93% of the time. This is a pretty good start at building a sentiment analyzer, with only a few lines of code (and a lot of data). You did not have to guess at the sentiment associated with a list of 7,500 words and hardcode them into an