← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8785.c · 生成模型:GPT-6

可扩展标注建立学习反馈

理解等级:中级

可扩展标注不是无差别逐条人工处理,而是利用用户信号、自然标签、专家判断或重点样本,把有限标注能力用在有价值的数据上。弱监督等方式能够辅助扩展,但噪声、偏差和适用范围需要验证;书中数量级提速是特定实践描述,不是普遍承诺。

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

      By scalable, we mean techniques where user interaction signals, expert judgments, natural labels, etc., can be applied to many data points, or to just the most important data points for model performance, rather than labeling every point individually and indiscriminately. This enables data-centric model improvements and enables teams to retrain their model regularly with better data, to keep up with a nonstationary world.

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第09章《MLOps与持续交付》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch09.html) - 结论依据:摘录中关于可扩展标注建立学习反馈的定义、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:By scalable, we mean techniques where user interaction signals, expert judgments, natural labels, etc., can be applied to many data points, or to just the most important data points for model performance, rather than labeling every point individually and indiscriminately. This enables data-centric model improvements and enables teams to retrain their model regularly with better data, to keep up with a nonstationary world.