Explore the intricacies of building a continual learning system for machine learning models in this comprehensive lecture. Delve into key concepts such as periodic retraining, iterative strategy development, and essential metrics for model monitoring. Learn about various retraining strategies, including logging, data curation, training triggers, dataset formation, and both offline and online testing methods. Gain insights into concrete continual learning workflows and take away valuable lessons for implementing these techniques in real-world scenarios. Access detailed notes and slides to supplement your learning experience and deepen your understanding of continual learning in machine learning systems.
Overview
Syllabus
Overview
How to think about continual learning
Periodic retraining
Iterating on your retraining strategy
Metrics for monitoring ML models
Tools and tests for metric monitoring
Retraining strategy: logging
Retraining strategy: data curation
Retraining strategy: training triggers
Retraining strategy: dataset formation
Retraining strategy: offline testing
Retraining strategy: online testing
A concrete continual learning workflow
Take-aways
Taught by
The Full Stack
