知识卡片
欠约束可能被同样测试分掩盖
多个配置可在同一测试数据上表现相近,却在部署环境不同,这是一种欠约束现象。随机种子敏感可能是线索,但不等于唯一诊断。因果识别是在特定模型和数据假设下查询是否唯一,不能直接与所有机器学习欠约束问题划等号。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 为何需要因果AI
原书出处 · 1 处
- · 查看原书依据
Underspecification occurs when there are many equivalent configurations of a model that perform equivalently on test data but perform differently in the deployment environment. One sign of underspecification is sensitivity to arbitrary elements of the model’s configuration, such as a random seed. Causal inference can tell you when a causal prediction is “identified” (i.e., not “underspecified”), meaning a unique answer exists given the assumptions and
参考来源
- 位置:《Causal AI》第01章《为何需要因果AI》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-1.xhtml)
- 结论依据:摘录中关于欠约束可能被同样测试分掩盖的定义、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:Underspecification occurs when there are many equivalent configurations of a model that perform equivalently on test data but perform differently in the deployment environment. One sign of underspecification is sensitivity to arbitrary elements of the model’s configuration, such as a random seed. Causal inference can tell you when a causal prediction is “identified” (i.e., not “underspecified”), meaning a unique answer exists given the assumptions and