知识卡片
迭代发布支持及时缓解伤害
小步与用户验证可更早发现不合适方案;可控发布和回退能力则在伤害暴露后缩短响应时间。CD4ML提供改善与纠错能力,但不能保证伤害从不发生,也不能靠事后回滚恢复所有已经造成的影响。
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
原书出处 · 1 处
- · 查看原书依据
Enabling iterative improvements CD4ML—the ability to make small changes to ML systems with a specific intent, evaluate and monitor if those changes had the desired result, and roll them back if not—is a powerful enabler for Responsible AI in multiple ways. First, it enables an iterative, human-centered approach to solution development, where we can research and test solutions with people in a series of small
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:摘录中关于迭代发布支持及时缓解伤害的定义、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:Enabling iterative improvements CD4ML—the ability to make small changes to ML systems with a specific intent, evaluate and monitor if those changes had the desired result, and roll them back if not—is a powerful enabler for Responsible AI in multiple ways. First, it enables an iterative, human-centered approach to solution development, where we can research and test solutions with people in a series of small