GPmp-contrib documentation¶
gpmp-contrib extends gpmp with computer-experiment objects, model
containers, Matérn container classes, sequential strategies, optimization
criteria, excursion/set-inversion tools, and relaxed Gaussian-process
utilities.
Use gpmp directly when you need models, kernels, numerical backends, or
parameter-selection functions. Use gpmp-contrib when you want computer
experiments, model containers, or sequential-design classes that coordinate
these objects.
For standard ModelContainer calls, model methods convert inputs by default
and prediction calls return NumPy arrays with convert_out=True. Explicit
backend conversion is needed when code passes arrays between gpmp internals,
posterior samplers, or external NumPy-only libraries.
Main documentation sections¶
- Installation
Package dependencies, backend notes, and documentation build commands.
- Getting started
Hartmann4 run: define a problem, build a model, select parameters, predict, and inspect diagnostics.
- User guide
Concepts and procedures: package organization, model containers, parameter objects, parameter selection, priors, diagnostics, sequential strategies, and reGP.
- Examples
Script-oriented explanations for the main examples in
examples/.- API reference
Public modules, classes, and functions.
- References
Literature cited by the guide and examples.
Main procedures¶
- First example
See Getting started. It builds the Hartmann4 example used in the core
gpmptutorial, selects Matérn covariance parameters, predicts at test points, and shows the expected diagnostic output.- Model construction
Read Models and computer experiments, Model state and parameter objects, and Parameter selection to choose a likelihood, restricted-likelihood, or REMAP parameter-selection rule and to understand stored parameter values.
- Priors in REMAP selection
Read Priors for REMAP selection when REMAP prior anchors or hyperparameters must be inspected or set by hand.
- Sequential design
Read Sequential design, optimization, and set estimation, then compare the fixed-grid and SMC examples.
- API details
Use API reference for signatures, shapes, return values, stored side effects, and failure conditions.