Deep Learning is Like Water (it's Everywhere!)
summary
Demonstration of a distributed H2O.ai stack that parses 116 million rows in 8 seconds across a 10-node cluster (≈320 cores), compresses 12GB to ~2GB, trains models (deep nets, GBM, XGBoost) in parallel, and exports production code as deployable POJO/Mojo Java packages. Live demos show GBM AUC ~0.746, XGBoost ~0.757, a simple 50/50 blend nudging AUC to ~0.749, GPU image training (MXNet/Caffe/TensorFlow backends) at ~37 samples/sec on tiny datasets that quickly overfit without larger data.
AI-written summary · about
chapters (7)
- 0:00Introduction to H2O
- 7:24Live distributed demo
- 15:26XGBoost integration demo
- 24:48Deep Water overview
- 28:41Image classification demo
- 35:21Multi-GPU training
- 39:59Future of AI and Q&A
related talks
42mH2O Deep Water with Python early sneek
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
44mDeep Learning in Production
34mDeep Learning at Scale
44mScalable Automatic Machine Learning with H2O
01h 18mIntroduction to H20 AutoML with Python