/B2BDC-PiecewiseModels

Piecewise surrogate models for Bound-to-Bound Data Collaboration toolbox

Primary LanguageMATLAB

B2BDC-PiecewiseModels

What are Piecewise Models for B2BDC?

  • Piecewise surrogate models is a class for B2BDC which constructs polynomial surrogate models on disjoint portions of a domain
  • Provides heuristic to subdivde domains
  • Subdomains which have large fitting error, but can still be proven incompatible with the experimental evidence, are neglected
  • Utilization of experimental evidence enables an increased efficiency, sampling regions where models and data agree
    • Number of evaluations of the underlying model can be reduced considerably compared to traditional piecewise surrogate modeling strategies

How does the class work?

The Piecewise models class is an extension to the Bound-to-Bound Data Collaboration toolkit. Methods and objects were tested with commit e9d501a

Installation

  • Copy piecewiseOptions.m into the root directory of the B2BDC Toolkit, i.e., +B2BDC/
  • Copy the folder @PiecewiseModel into the directory +B2BDC/+B2Bmodels/

Methods

findModelIndex
eval
length
rule
inDomain
grow
extractData
fitSubDomain
consistentDomains
branch