cran / modelcf Goto Github PK

Author : Benjamin Auder
Institution : CEA Cadarache, UPMC Paris 6

This package contains a generic way to build surrogate models of physical computer codes, when inputs are vectors (in R^p) and outputs (continuous) curves from [a,b] to R. The curves are discretized on a finite grid t1,...,tD.

-------------------------------------------

Main function to build a model = fmetam

-------------------------------------------

Main function to validate a model = nfoldcv


*******************************************

Main improvements over 2.0 version :
  - several bug fixes, especially for dimensionality reduction
  - several mehods added for intrinsic dimensionality estimation
  - possibility to better fit to the data by learning residuals


*******************************************

Acknowledgements :
This work was supported by the CEA Cadarache. Thanks to Bertrand Iooss, Michel Marques and all the people who helped me - directly or not - to improve the program design.