API reference
Core
The HistoryMatching engine and the domain objects it operates on.
| engine | HistoryMatching — the main user-facing class for running a calibration. |
| parameter_space | ParameterSpace domain object for history matching. |
| observation_data | ObservationData domain object for history matching. |
| emulator_bank | EmulatorBank domain object for history matching. |
| iteration_result | IterationResult — the result of a single history matching wave. |
Strategies
Pluggable components for drawing samples, choosing which outputs to emulate, and sampling the NROY region.
| sampling | Sampling strategy implementations for history matching. |
| feature_selection | Feature selection strategy implementations for history matching. |
| nroy_sampling | NROY sampling pipeline inspired by the hmer R package. |
Emulators
Emulator implementations and the factory that creates them. BayesLinear is the default; subclass BaseEmulator and register it with EmulatorFactory to add your own.
| emulators.base | Abstract base class that every history matching emulator implements. |
| emulators.bayes_linear | Bayes Linear emulator inspired by the hmer R package. |
| emulators.gpr | Gaussian Process Regression (GPR) emulator. |
| emulators.glm | Generalized Linear Model (GLM) emulator. |
| emulators.linear | Ordinary least squares (OLS) linear regression emulator. |
| emulators.factory | Emulator factory with strategy pattern for history matching. |
| emulators.results | EmulationResults class for standardized emulator output handling. |
| emulators.fast_predict | Fast GPR prediction using numba, with no TensorFlow call at prediction time. |
Plotting
Composable plotting helpers. Each returns Matplotlib axes and accepts an ax=/axes= argument; the most commonly used are re-exported at the top level, e.g. historymatching.plot_pairplot.
| plotting | Reusable plotting helpers for history matching. |