Auto ML
summary
Claims that AutoML speeds prototyping, demonstrated by using H2O AutoML to build a distortion estimator for industrial metal 3D printing, producing a working prototype in three days that is accurate enough for engineers and ready to deploy. Workflow relies on heavy feature engineering and telemetry (temperature, vibration, pressures), shows AutoML handles model search so teams avoid manual hyperparameter tuning, but warns of limited model repertoire and extreme sensitivity to preprocessing — AutoML speeds iteration but does not replace ML expertise.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:02Why this talk matters
- 3:10ML project challenges
- 6:23Fail fast succeed faster
- 9:33AI complexity vs simplicity
- 12:27Mitigating AI uncertainty
- 14:32What is AutoML
- 16:39AutoML frameworks
- 18:27First steps in ML
- 21:17Managing AI expectations
- 24:07AutoML for prototyping
- 26:36Unknowns in AI projects
- 28:07AutoML limitations
- 30:15People management with AutoML
- 32:03Case study: 3D printing distortion
- 36:18Results and effectiveness
- 37:47Q&A
related talks
08mBroadening AI Adoption with AutoML
28mAutoML—hype vs. value
21mDemocratizing AI: AutoML Transforms Predictive Analytics
01h 08mAutoML: 40 hours of doing machine learning without knowing it
28mHow to boost data science routines with DataArt’s AutoML Platform?
33mML: Automated Machine Learning
44mScalable Automatic Machine Learning with H2O
01h 18mIntroduction to H20 AutoML with Python