tgp: Bayesian Treed Gaussian Process Models for Nonlinear Regression

An R package implementing Bayesian treed Gaussian process models for nonstationary, semiparametric nonlinear regression and experimental design. The software includes special cases like Bayesian linear models, CART, and stationary GPs, and provides functions for visualization, sensitivity analysis, and sequential design. It was authored by Robert B. Gramacy, with methodology detailed in the Journal of Statistical Software in 2007 and 2010.

Use Cases

• Fit nonstationary, semiparametric nonlinear regression models based on the treed Gaussian process framework.
• Perform sequential experimental design and adaptive sampling using provided functions like ALM, ALC, and expected improvement.
• Conduct sensitivity analysis and build multi-resolution models as supported by the package.
• Visualize model output using the provided 1-d and 2-d plotting and tree drawing functions.

Strengths

• Implements a flexible modeling framework that includes several special cases like Bayesian linear models and CART.
• Provides dedicated functions for visualization, sensitivity analysis, and sequential experimental design.
• Methodology is peer-reviewed and documented in the Journal of Statistical Software.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

paperswithcode

Collection Method

Software package release.