← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6681.b · 生成模型:GPT-6

主题先验影响混合稀疏程度

理解等级:中级

Dirichlet的参数描述主题比例的倾向,不只是主题数的另一种写法。不同参数可让文档更集中于少数主题或更均匀;实现是否同时给主题词分布配置先验也须查明,不能把词频本身直接叫Dirichlet分布。

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

      All this machine needs is a single parameter for that Poisson distribution (in the dice roll from step 1) that tells it what the average document length should be and a couple more parameters to define that Dirichlet distribution that sets up the number of topics. Then, your document generation algorithm needs a term-topic matrix of all the words and topics it likes to use,

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第04章《从词数到语义分析》(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-4.html) - 结论依据:摘录中关于主题先验影响混合稀疏程度的定义、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:All this machine needs is a single parameter for that Poisson distribution (in the dice roll from step 1) that tells it what the average document length should be and a couple more parameters to define that Dirichlet distribution that sets up the number of topics. Then, your document generation algorithm needs a term-topic matrix of all the words and topics it likes to use,