From the course: Data-Centric AI: Best Practices, Responsible AI, and More
Join today to access over 24,900 courses taught by industry experts.
Monitoring and maintaining ML models in production
From the course: Data-Centric AI: Best Practices, Responsible AI, and More
Monitoring and maintaining ML models in production
- [Instructor] As we've discussed, data-centric AI requires continuous monitoring and adaptation to changes in real-world data. This brings us to the key topics of data drift and model drift. We'll explore different forms of drift and how they degrade model performance over time. I'll also provide guidance on how to monitor models in production to detect these issues. We'll dive into data drift, specifically when data drift distribution shifts from the original training data. As we'll discuss model drift referring to changes in the relationship between the variables, you'll gain theoretical grounding on how and why these drifts occur. You'll also cover leading techniques to detect drift using statistical tests and other monitoring strategies. Understanding drift empowers us to take corrective actions, like retraining on updated data and tuning model hyperparameters. This knowledge enables resilient and adaptable AI…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.