Recommended
Recommended
More Related Content
What's hot
What's hot (20)
Similar to Big Data LDN 2017: H2O.ai Driverless AI: Fast, Accurate, Interpretable AI
Similar to Big Data LDN 2017: H2O.ai Driverless AI: Fast, Accurate, Interpretable AI (20)
More from Matt Stubbs
More from Matt Stubbs (20)
Recently uploaded
Recently uploaded (20)
Big Data LDN 2017: H2O.ai Driverless AI: Fast, Accurate, Interpretable AI
- 1. H2O.ai Driverless AI Using AI to do Fast, Accurate and Interpretable AI Jo-fai (Joe) Chow Data Scientist joe@h2o.ai @matlabulous Big Data LDN 15th November, 2017
- 2. Agenda • Introduction • Company / People • Using AI to do AI • Evolution of H2O Platform • Motivation • Driverless AI Key Features • Credit Risk Demo • Other News 2
- 3. Company Overview Founded 2011 Venture-backed, debuted in 2012 Products • H2O Open Source In-Memory AI Prediction Engine • Sparkling Water (H2O + Spark) • Enterprise Steam • Driverless AI Mission Operationalize Data Science, and provide a platform for users to build beautiful data products Team 75+ employees • Distributed Systems Engineers doing Machine Learning • World-class visualization designers Headquarters Mountain View, CA 3
- 6. 6 H2O Team
- 7. Using AI to do AI H2O.ai Driverless AI 7
- 8. H2O AI Platform Timeline 8 Users Advanced Data Scientists Developers/Engineers Dev Ops App Developers H2O Core Data.table Sparkling Water Roadmap Steam Visual Interpretation Deep Water Analysts H2O Driverless AI 2012 2014 2016 2017 2018 H2O GPU Edition GPU 2019 Auto ML
- 9. “Confidential and property of H2O.ai. All rights reserved” Shortage of Data Scientists “The United States alone faces a shortage of 140,000 to 190,000 people with analytical expertise and 1.5 million managers and analysts” –McKinsey Prediction for 2018
- 10. Kaggle Masters at H2O.ai 10
- 11. The Product: Driverless AI • Kaggle Grandmasters in a Box • A solution to the shortage of data scientists • Fast • H2O Algorithms optimised for GPUs • Accurate • Auto Feature Engineering • Auto Model Tuning / Selection / Ensemble • Interpretable • Machine Learning Interpretation • Production Ready • Export pipeline as a standalone Python package 11
- 12. Fast: H2O Algorithms on GPUs 12 3400+ models were trained on GPUs by the time the first model completed training on CPUs https://www.youtube.com/watch?v=LrC3mBNG7WU
- 13. Accurate: Auto Feature Engineering 13 Feature Transformation Pipeline Model Training Feature Engineering
- 14. Accurate: Auto Feature Engineering 14 • Cross Validation Categorical Encoding • Frequency Encoding • Cross Validation Target Encoding • Truncated SVD and More Feature Transformations Generated Features Original Features
- 15. Accurate: Auto Model Tuning/Selection/Ensemble 15
- 16. Interpretable: Explainable Machine Learning 16 https://www.oreilly.com/ideas/ideas-on-interpreting-machine-learning
- 17. Demo: Credit Card Default Risk 17
- 18. 18
- 20. 20 Auto Feature Engineering Trying out Different Sets of Features Using both CPUs & GPUs
- 21. 21 Options for Interpretation, Scoring, Logs, Pipeline
- 22. 22
- 23. 23 Zoom into individual records Explain how inputs lead to outcome with a simple decision tree
- 24. 24 Turn decision tree into reason codes
- 29. Using AI to do AI H2O.ai Driverless AI 29
- 30. Other News 30
- 32. Thank you everyone for your support! 32 4000+ Members 450+ RSVPs!!!
- 33. London AI & Deep Learning Meetup 33 15th November, 2017 Keynote Theatre, Olympia - London Time Session Title Speaker 6:00 – 6:30 pm Doors Open Pizzas, Drinks & Networking / 6:30 – 6:45 pm Introduction Joe 6:45 – 7:30 pm From Kagglenoobs to Kaggle Master Yifan 7:30 – 7:45 pm Break / 7:45 – 8:30 pm Scaling Machine Learning with H2O Sparkling Water Kuba 8:30 – 9:00 pm Q&A and Networking / Sponsored by
- 34. • Organisers 34 Thank you! • Code, Slides & Documents • bit.ly/h2o_meetups • docs.h2o.ai • Contact • joe@h2o.ai • @matlabulous • github.com/woobe • Please search/ask questions on Stack Overflow • Use the tag `h2o` (not H2 zero)