H2O Feature Store to Driverless AI Workflow
- Predictive AI
- H2O Feature Store
- H2O Driverless AI
This lesson tours the H2O Feature Store and H2O Driverless AI as connected pieces of one machine learning platform workflow. It shows how to browse public and private projects and feature sets in the Feature Store, including versioning that lets you retrieve a feature set's latest state or a specific point in time. The presenter then registers a data set from a notebook, spins up an AI engine sized automatically for the workload, and imports the data into Driverless AI. The lesson closes inside Driverless AI, showing a credit default project's training and test splits with auto visualization.
Key takeaways
- How the Feature Store tracks multiple versions of a data set so you can retrieve the latest version or a specific point in time
- Why AI engines can be sized automatically based on data size and task type, with auto timeout after an hour of idle use
- How to move a registered data set from the Feature Store into an H2O Driverless AI project using Python code
- What auto viz does inside Driverless AI: automatically creates data set visualizations to help you understand your data
- How access controls let you keep a Feature Store project secret and shared with a specific group instead of the full organization
Start owning your AI