Moving from Offline to Online Machine Learning with River
summary
River brings online ML to Python by training models on streaming data one event at a time and updating state incrementally, which cuts memory/CPU compared with batch retraining and learns one observation at a time. It ships models and features as dictionaries so you can inspect and dump models as JSON, supports pipelines with scalers, anomaly detection via half‑space trees, simulates delayed labels with progressive validation, and integrates into Kafka stream processors for low‑latency inference and adaptive handling of concept drift.
AI-written summary · about
chapters (12)
- 0:00Introduction to real-time ML
- 1:24Stream processing background
- 5:56Offline vs online ML
- 8:47Conceptual data flow
- 10:31Online inference explained
- 11:36Online learning fundamentals
- 13:00Benefits of online ML
- 14:23Use cases and drift resilience
- 15:28Batch and stream complementarity
- 15:48Introducing River library
- 19:39Learning paths with River
- 20:21Q&A session
related talks
30mMoving from Offline to Online Machine Learning with River
31mMoving from Offline to Online Machine Learning with River
30mReconceptualizing Machine Learning for the Real-time World
41mStreams to the River: Powering an Online XGBoost Classifier with Kafka
22mOnline Machine Learning with Creme
37mAutomated Machine Learning: Building a Pipeline
31mRunning Machine Learning Analytics On Traces
29mBatch and Streaming Processing in the World of Data Engineering and Data Science