原书出处
Reliable Machine Learning
09 Monitoring and Observability for Models
Test in production, with the model getting a small subset of real user traffic (typically, between 1% and 5%), sometimes called a canary test. The second approach also has the good effect of exposing any problem with the model reacting to user traffic before a full rollout.
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch09.xhtml
EPUB SHA-256:1201425671b4d9d06d56459bced6f75ec74a7aac380ea24f5b573c80fee7be65