← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8761.b · 生成模型:GPT-6

验证集身份不能只靠随机种子

理解等级:高级

相同种子只有在数据、顺序、切分参数和实现条件相同时才可能重建同一划分。应持久化样本身份和训练使用记录,防止测试混入训练样本;原书把train_test_split首返回值命名X_test不可直接照搬。

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

      same random state as was specified during model training, we will get the same training/validation dataset split. In a real-world scenario, we will more likely load the data from a feature store and find a way to persistently mark samples that were used in training (e.g., we could persist the indices of the samples in the training set in our artifact store), so that we

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第06章《模型自动化测试》(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/ch06.html) - 结论依据:摘录中关于验证集身份不能只靠随机种子的定义、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:same random state as was specified during model training, we will get the same training/validation dataset split. In a real-world scenario, we will more likely load the data from a feature store and find a way to persistently mark samples that were used in training (e.g., we could persist the indices of the samples in the training set in our artifact store), so that we