User Reference Guide

pyapprox provides functions for surrogate modeling, sensitivity analysis, quadrature, inference, optimal experimental design and multi-fidelity modeling

Surrogate modeling

Supervised learning

The pyapprox.approximate.approximate() function produces response surface approximations from training data.

The approximate function supports the following methods

Supervised active learning

The pyapprox.approximate.adaptive_approximate() function produces adaptive response surface approximations of a function fun, As the approximation is built the function being approximated is sampled at locations that greedily minimize an estimate of error.

The adaptive_approximate function supports the following methods

Sensitivity analysis

Surrogate based global sensitivity analysis

The following functions can be used to extract sensitivity metrics analytically from a surrogate.

The following functions can be used to visualize variance based sensitivity measures

Local sensitivity analysis

The pyapprox.sensitivity_analysis.analyze_sensitivity_morris() computes morris sensitivity indices.

Multivariate Quadrature

Surrogate based quadrature

The following functions can be used to compute the mean and variance analytically from a surrogate.

Inference

The function pyapprox.bayesian_inference.markov_chain_monte_carlo.run_bayesian_inference_gaussian_error_model() can be used to draw samples from the posterior distribution of variables of a model conditioned on a set of observations with Gaussian noise.

Optimal experimental design

Optimal experimental designs for m-estimators such as least squares and quantile regression can be computed with

pyapprox.optimal_experimental_design.optimal_experimental_design()