今天就开通帐号，24,600 门业界名师课程任您挑！
Introduction to Lakehouse Monitoring
- [Narrator] Monitoring is one of the biggest challenges in the MLOps world. There are a lot of things to monitor. As for any software product, we need to monitor application and infrastructure health, application performance and costs. On top of that, machine learning application must be monitored for model and data drift, and the business value it delivers. Some machine learning models, for example, fraud detection or loan applications must be also monitored for fairness and bias. Data drift occurs when the distribution of the input data changes over time. For example, in the hotel booking cancellation use case, we use the number of adults and the number of children to predict the likelihood of canceled booking. Let's say the system started offering special discounts for families changing the distribution of these variables. User preferences did not change, nor the relationship between the input and the outcome. This data drift may have resulted in the overall drop in model…
下载课程离线学习
下载课程到移动设备，轻松离线学习。在领英学习 iOS 或安卓版 APP 下载课程