Active Learning Examples

This section collects the active-learning tutorials that demonstrate JetGP’s adaptive design-of-experiments framework. The theory underlying these examples is described in Cost-Aware Active Learning with Directional Derivatives.

• Cost-Aware Active Learning on the 2D Branin-Hoo Function
  • Overview
  • Step 1: Define the Branin-Hoo Function
  • Step 2: Configure the Active-Learning Run
  • Step 3: Inspect the Sequential Design
  • Step 4: Inspect the Final Surrogate
  • Reproducing the Figures
  • Summary
• Budgeted Cost-Aware Policy Comparison
  • Overview
  • Baseline Policies
  • Experiment 1: Derivatives Are Cheap
  • Experiment 2: Function Values Are Cheap
  • Results
  • Interpretation
  • Reproducing the Figure
• Naive Gradient-Enhanced Baseline vs. Cost-Aware Active Learning
  • Overview
  • Cost Accounting
  • Methods
  • Result
  • Why the gap is so large
  • Takeaways
  • Reproducing the Figure