Skip to content

Tutorials

These tutorials walk through how to use the history matching library, from basic setup to advanced workflows. To run locally, start a Jupyter environment in the docs/tutorials folder.

Getting started

  • Basic workflow — Complete walkthrough calibrating a model with the single-constructor HistoryMatching API
  • Manual workflow — Fine-grained control over each iteration with manual inspection and decision points

During calibration

  • Advanced configuration — Custom strategies, strategy switching mid-workflow, checkpoint/resume
  • Emulator showcase — Comparison of Bayes linear, linear, GLM, and GPR emulators; a key takeaway is that Bayes linear matches GPR quality at a fraction of the cost (hence the default)
  • NROY sampling methods — Comparison of sampling strategies for exploring the non-implausible space
  • Visualization — Composable plotting helpers for NROY corner plots, convergence, marginals, and emulator quality

After calibration

  • Trajectory selection — Draw plausible (θ, seed) pairs (here θ = (β, γ)) for stochastic simulation using importance resampling