← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6557.c · 生成模型:GPT-6

训练监控集不能兼任最终验收

理解等级:中级

反复查看某个留出集以调架构、学习率或停止时间,会使它承担验证集职责。若还要报告独立泛化结果,应保留未参与选择的最终测试集。随机重建图可作诊断,但不能代替系统的样本外评价。

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

      test data to evaluate how well training is progressing. You can include anything you think is useful to monitor during training. Here, I calculate and print the loss function on the test data, just to make sure the test loss is progressively decreasing along with training loss (a flattening of test loss while training loss continues to decrease would indicate overfitting).

    参考来源

    - 位置:《Causal AI》第05章《连接因果与深度学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-5.xhtml) - 结论依据:摘录中关于训练监控集不能兼任最终验收的定义、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:test data to evaluate how well training is progressing. You can include anything you think is useful to monitor during training. Here, I calculate and print the loss function on the test data, just to make sure the test loss is progressively decreasing along with training loss (a flattening of test loss while training loss continues to decrease would indicate overfitting).