原书出处
Reliable Machine Learning
15 Case Studies: MLOps in Practice
Because the corrupted data included a lot of traffic with no associated clicks, the model was getting a signal that the overall click-through rate in the world was now about half of what it had been just a few hours before. The resulting learning within the model soon led to changes in the served models and, not surprisingly, lower served predictions than normal.
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch15.xhtml
EPUB SHA-256:1201425671b4d9d06d56459bced6f75ec74a7aac380ea24f5b573c80fee7be65