SACOBRA: Self-adjusting Constrained Optimization with RBF

output_Op3uU6

Package SACOBRA can be downloaded from this CRAN-URL. SACOBRA is a derivative-free optimizer which is able to solve constrained expensive problems with very few function evalutions. This package works based on surrogate assisted techniques and utilizes RBF interpolation to model fitness function and constraint functions.

 


An R-interface to the JAVA version of CMA-ES by Niko Hansen (rCMA)

Package rCMA can be downloaded from this CRAN-URL.CMA_cartoon

rCMA is a package to perform CMA-ES optimization, using the Java implementation by Niko
Hansen [Hansen,2009]. CMA-ES [Hansen and Ostermeier,1996], [Hansen,2013] is
the Covariance Matrix Adapting Evolutionary Strategy for numeric black box optimization.

Read more:

 

 

 


R-Package for Sequential Parameter Optimization Toolbox (SPOT)

An R version of this toolbox for interactive and automatic optimization of algorithms can be downloaded from this CRAN-URL. spotOfficial

 

 

 

 

 



 

R-Package for Tuned Data Mining (TDMR)

Important note: Due to a major new SPOT release 2.0, TDMR is currently not available on CRAN.
To download and install TDMR, first download
       SPOT_1.1.0.tar.gz
       TDMR_1.5.tar.gz
and then issue from R these commands
       install.packages("SPOT_1.1.0.tar.gz", repos = NULL, type = "source")
       install.packages("TDMR_1.5.tar.gz", repos = NULL, type = "source")
while being in the directory of the .tar.gz files.

(If desired, the older version TDMR_1.4.tar.gz is available as well.)

28921_small_christophe-papke_pixelio.de_
An R version of this toolbox for interactive and automatic tuning of data mining tasks can be downloaded from this CRAN-URL. 28921_small_christophe-papke_pixelio.de_
Read more:
  • TDMR project page
  • TDMR tutorial (CIOP-Report 03/2012 [Kone12b], last update May 2016)
  • TDMR-docu.pdf (CIOP-Report [Kone12a], last update June 2017): User manual, for in-depth information on usage and development of the TDMR package.

 

 

 


Slow Feature Analysis Toolkit SFA-TK

 

Slow Feature Analysis (SFA) is a technique developed by Laurenz Wiskott to find features in complex timeseries or multivariate datasets. SFA-TK V2.6 – V2.8 by Wolfgang Konen are extended versions of the original MATLAB-based SFA-Toolkit (SFA-TK V1.0 by Pietro Berkes).

SFA-TK-V1.0

 

 

 

 

 

SFA-TK V2.8 has in addition to V2.7: parametric bootstrap, regularization of Gaussian classifier, nearest neigbor classifier

SFA-TK V2.7 has in addition to V2.6:  storable and loadable classifier models, demo dataset descriptions.

SFA-TK V2.6: The extensions are: SVD-based SFA, support for classification, Gaussian classifier, some classification demo datasets (UCI, gesture) with or w/o cross validation.

Download SFA-TK from the following URLs:

For developers: An algorithmic-mathematical summary of the SFA procedure for classification is available here (in German only). For more literature on SFA see our SFA page.