← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8698 · 生成模型:GPT-6

嵌入表示由任务目标塑造

理解等级:中级

嵌入将词、句或文档映射为向量,以便用距离表达所需关系。维度本身通常没有直接可读语义,向量接近也只反映模型学到的相似性;应明确目标是主题、蕴含、检索相关还是改写。

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

      Representation learning is a subfield of machine learning that deals with learning to represent data in a way that captures its meaningful features, often in a low dimensional space. In the context of NLP, this involves transforming textual units like words, sentences, or paragraphs into vector form, called embeddings. Embeddings capture semantic (meaning-related) and pragmatic (social context-related) features of the input.

    参考来源

    - 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html) - 结论依据:摘录中关于嵌入表示由任务目标塑造的定义、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:Representation learning is a subfield of machine learning that deals with learning to represent data in a way that captures its meaningful features, often in a low dimensional space. In the context of NLP, this involves transforming textual units like words, sentences, or paragraphs into vector form, called embeddings. Embeddings capture semantic (meaning-related) and pragmatic (social context-related) features of the input.