知识卡片

普通读书笔记卡 · 7009.g · 生成模型:DeepSeek-V4.1-Flash

特征的定义

理解等级:初级

训练数据是我们认为与模型相关的、关于世界中事件或事实的数据;特征(feature)则是该数据中一个具体的、可测的方面,尤其是我们判断最有助于在相似情境下建模、分类与预测未来事件的那些方面。特征开发也是实验的主要抓手:做模型的过程很大程度上就是弄清数据的哪些方面对任务有用、哪些无关,因此需要能轻松新增特征、在旧数据上按新想法重算特征(回填)、以及移除被证明不重要的特征。这也解释了为什么特征存储的可追溯性与一致性直接决定团队迭代速度,而不只是存储格式问题。

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

    • Training Systems · 查看原书依据

      A feature is a specific, measurable aspect of that data. Specifically, features are those aspects of the data that we believe are most likely to be useful in modeling, categorizing, and predicting future events given similar circumstances. To be useful, features need a consistent definition, consistent normalization, and consistent semantics across the whole ML system, including the feature store, training, quality evaluation,

    参考来源

    - 位置:《Reliable Machine Learning》第07章《Training Systems》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch07.xhtml) - 结论依据:原文明确阐述该主题,并给出定义、实践或反例。 - 原始内容:A feature is a specific, measurable aspect of that data. Specifically, features are those aspects of the data that we believe are most likely to be useful in modeling, categorizing, and predicting future events given similar circumstances. To be useful, features need a consistent definition, consistent normalization, and consistent semantics across the whole ML system, including the feature store, training, quality evaluation,