Reading the Driverless AI Experiment Page
- Observability
- H2O Driverless AI
This lesson explains what to watch on the Driverless AI experiment page while an experiment runs through data preparation, validation splits, and model and feature tuning. It covers why the experiment setup and training settings lock once launched, and what the CPU and memory section's insights, scores, notifications, logs, and trace views show along the way. It focuses mainly on the iteration graph, where each green dot is a model trained that iteration, hovering reveals its type and feature count, and the minimum and maximum scorer values show the experiment's performance range.
Key takeaways
- Why the experiment setup and training settings lock once an experiment launches, requiring a new experiment to change them
- What the CPU and memory section's insights, scores, notifications, logs, and trace views each show during a run
- How each green dot on the iteration graph represents one model trained with a specific number of features
- How hovering over a dot on the iteration graph reveals that model's type and the features it was trained with
- What the minimum and maximum scorer values on the iteration graph tell you about the experiment's performance range
Start owning your AI