SACOBRA (version 1.2)

Self-Adjusting COBRA

Description

Performs surrogate-assisted optimization for expensive black-box constrained problems.

Copy Link

Version

Install

install.packages('SACOBRA')

Monthly Downloads

Version

1.2

License

GPL (>= 2)

Maintainer

Samineh Bagheri

Last Published

March 26th, 2020

Functions in SACOBRA (1.2)

cobraPhaseI

Find a feasible solution.

debugVisualizeRBF

Optional visualization of surrogate models for 2-dimensional problems, d=2.

defaultCA

Default settings for online whitening functionality

cobraPhaseII

Improve the feasible solution by searching new infill points

DRCS

Distance Requirement Cycle, short version

RandomStart

Random start Algorithm

cobraInit

Initial phase for SACOBRA optimizer

defaultDebugRBF

Default settings for debug visualization RBF (only for d==2)

SACOBRA-package

Self-adjusting Constrained Optimization with RBF Surrogates

DRCL

Distance Requirement Cycle, long version

COP

Constraint Optimization Problem Benchmark (G Function Suite)

forwardRescale

Forward Rescaling

fnArchiveFactory

construct wrapper fnArchiveF for a function fn

defaultEquMu

Default settings for equality handling mechanism

multiRunPlot

Plot the results from multipe COBRA runs.

trainCubicRBF

Fit cubic RBF interpolation to training data X for d>1.

multiCOBRA

Perform multiple COBRA runs

trustRegion

Performs trust region refinement

trainGaussRBF

Fit Gaussian RBF model to training data for d>1.

defaultMS

Default settings for the model-selection part of SACOBRA.

distLine

Euclidean distance of x to all xp

getFbest

Return best objective function value

defaultRI

Default settings for repairInfeasRI2 and repairChootinan.

defaultSAC

Default settings for the SACOBRA part of SACOBRA.

multiRunPlot_2

Plot the results from multipe COBRA runs.

evalReal

Evaluate new iterate on real function(s)

repairInfeasibleW

Wrapper for repairInfeasRI2 (needed by RBFsearch.R).

rescaleWrapper

Return a rescaled function

getXbest

Return best feasible solution in original space

plog

Monotonic transform

updateCobraEqu

Calculate cobra$xbest,fbest,ibest given the currently evaluated points.

trainMQRBF

Fit multiquadric RBF model to training data for d>1.

repairInfeasRI2

Repair an infeasible solution with the method RI2

repairChootinan

Repair an infeasible solution with the method of Chootinan.

defaultTR

Default settings for the trust-region part of COBRA.

interpRBF

Apply the trained cubic, MQ or Gaussian RBF interpolation to new data for d>1.

inverseRescale

Inverse Rescaling

trainSurrogates

Training surrogates for objective and constraint functions

intern.archive.env

Archiving Environment

setOpts

Merge the options from a partial list and the default list

startCobra

Start COBRA (constraint-based optimization) phase I and/or phase II

modifyMu

Modify equality margin $\mu$

plogReverse

Inverse of plog

updateSaveCobra

Update and save cobra

predict.RBFinter

Apply cubic or Gaussian or MQ RBF interpolation