Scalable Automatic Machine Learning with H2O
summary
H2O AutoML runs distributed, Java‑based algorithms via Python/R/Scala, performs automatic target encoding for high‑cardinality categoricals (example: ~40,000 ZIP codes), uses random search plus per‑model tuning with automatic early stopping, and builds stacked ensembles: one with all models and a best‑of‑family ensemble (one top model per family). XGBoost is bundled and can run on GPUs; models export as Java for production. On a 100‑minute Kaggle run it scores 8th, showing the pipeline works.
AI-written summary · about
chapters (11)
- 0:00Introduction to scalable AutoML
- 1:55What is H2O and its features
- 6:35Defining automatic ML
- 11:10Data preprocessing in AutoML
- 15:43AutoML techniques comparison
- 18:37H2O AutoML implementation
- 22:14Running H2O AutoML in Python and R
- 23:22Interpreting AutoML leaderboard
- 25:11AutoML on Kaggle
- 26:57Q&A: Spark ML vs H2O and GPU
- 29:03Future of AutoML
related talks
44mScalable Automatic Machine Learning with H2O
01h 18mIntroduction to H20 AutoML with Python
50mScalable Data Science & Deep Learning with H2O
01h 10mIntroduction to Machine Learning with H2O and Python
27mDeploying distributed AI and machine learning in financial services
01h 26mMachine Learning Algorithms with H2o and PowerShell
01h 10mGetting started with H2O on Python
20mMulti-algorithm Ensemble Learning at Scale: Software...