Getting Started with H2O Driverless AI
- Predictive AI
- H2O Driverless AI
Overview
This webinar, hosted by H2O.ai's marketing and partnerships engineering teams, gives a high-level tour of Driverless AI alongside H2O's open-source H2O-3 and Sparkling Water platforms. It walks through Driverless AI's seven core components (Datasets, Projects, AutoViz, Experiments, Diagnostics, MLI, and Deployments) before a live demo that loads a credit card default dataset, runs an AutoML experiment, reviews model interpretability, and deploys the resulting model as a REST server. It closes with audience Q&A covering handling of text and OCR data.
Key takeaways
- The difference between H2O's products: H2O-3 and Sparkling Water (open-source, library-style) versus Driverless AI (the enterprise AutoML platform)
- What each of Driverless AI's seven components does: Datasets, Projects, AutoViz, Experiments, Diagnostics, MLI (machine learning interpretability), and Deployments
- How to load a dataset, split it into train and test sets, and use AutoViz to spot correlations and potential leakage before running an experiment
- How the accuracy, time, and interpretability knobs in an experiment control search space, runtime, and how explainable the resulting feature transformations are
- How MLI surrogate models (like decision trees and linear models), Shapley plots, and disparate impact analysis help explain what a Driverless AI model is doing and why, which matters for regulated use cases
- How to one-click deploy a trained model as an AWS Lambda function or a local REST server for testing
Start owning your AI