知识卡片
数据即代码:训练数据决定系统行为
训练数据是定义系统行为的基础元素:数据的质量塑造了模型与系统的质量与行为,其中的不完美还会以出人意料的方式被放大。从这个视角看,数据在 ML 系统中的角色可类比传统系统里的代码。但传统代码能用前置条件、后置条件与不变量描述、也往往可系统测试;真实数据却带有有机的缺陷与不规则,无法如此规整刻画。因此数据问题会演化出覆盖不全、伪相关、冷启动、反馈回路、世界变化等多种故障,且不像代码 bug 有清晰报错,必须从数据本身入手发现与防御。
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
- · 训练数据的脆弱性
原书出处 · 1 处
- Basic Introduction to Models · 查看原书依据
From this lens, the role of data in ML systems can be seen as analogous to code in traditional systems. But unlike traditional code, which can be described with preconditions, post-conditions, and invariants, and may also be systematically testable, real-world data typically has organic flaws and irregularities. These can lead to a variety of problem types.
参考来源
- 位置:《Reliable Machine Learning》第03章《Basic Introduction to Models》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch03.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:From this lens, the role of data in ML systems can be seen as analogous to code in traditional systems. But unlike traditional code, which can be described with preconditions, post-conditions, and invariants, and may also be systematically testable, real-world data typically has organic flaws and irregularities. These can lead to a variety of problem types.