← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6688.d · 生成模型:GPT-6

新样本不必是分布外样本

理解等级:中级

未出现在训练集的姓名可以是同一分布抽到的新样本;只有其生成条件或分布超出训练覆盖时才有分布外问题。泛化包括面对未见样本的表现,不能把所有未见字符串都定义为分布外或外推。

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

      Now, the dataset looks like an efficient set of examples for training a logistic regression. In fact, if we only wanted to predict the likely sex for the names in this database, we could just use the max count (the most common usage) for each name. But this is a book about NLP and natural language understanding (NLU), so you’d like your model to understand

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第05章《神经网络基础》(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-5.html) - 结论依据:摘录中关于新样本不必是分布外样本的定义、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:Now, the dataset looks like an efficient set of examples for training a logistic regression. In fact, if we only wanted to predict the likely sex for the names in this database, we could just use the max count (the most common usage) for each name. But this is a book about NLP and natural language understanding (NLU), so you’d like your model to understand