TUHH > Servicebereiche > Rechenzentrum > Intlab

INTLAB - INTerval LABoratory

Neuigkeiten:

23.1.12: Im Studi-Pool ist testweise Intlab_V6 unter Matlab R2011b installiert worden. In startup.m sollte stehen:

path('/progsys/matlab/2011b/toolbox/Intlab_V6', path)
startintlab
Danach einfach matlab aufrufen.

3.7.09: Im Linux-Pool ist testweise Intlab_V5.5 unter Matlab R2009a installiert worden. In startup.m sollte stehen:

path('/progsys/matlab/2009a_crk/toolbox/Intlab_V5.5', path)
startintlab
Danach einfach matlab2009a aufrufen.

INTLAB ist die von S. Rump für Matlab entwickelte Toolbox für selbst-validierende Algorthmen. Sie steht in den Linux-Pools für Matlab R2006a/b in der Version 5.3 zur Verfügung.

Nutzbar ist INTLAB durch folgende Einträge ins startup.m der eigenen Matlab-Ressourcen:

   system('printenv BLAS_VERSION');
   path('/progsys/matlab/2006b/toolbox/Intlab_V5.3', path)
   startintlab

Vor dem Aufruf matlabs ist die Umgebungsvariable zur Wahl der BLAS-Routinen zu setzen:

setenv BLAS_VERSION atlas_P4.so

Danach sollten nach dem Start matlabs (matlab2006b) folgende Meldungen erscheinen:

===> Initialization of INTLAB toolbox
===> rounding checked and no errors detected
 
**************************************************************
*** Welcome to INTLAB - INTerval LABoratory Version 5.3    ***
***   Siegfried M. Rump, Insitute for Realiable Computing  ***
***   Hamburg University of Technology, Germany            ***
**************************************************************
 
===> rounding checked and no errors detected
===> width of display 120 characters
===> Default display of intervals with uncertainty (e.g. 3.14_) 
===> Rigorous standard functions in use
===> Complex interval stdfct used automatically for real interval input 
         out of range, but with warning
===> Improved residual calculation in verifylss
===> Verifylss for dense systems applies second stage only if first stage fails
===> Fast interval matrix multiplication in use (maximum overestimation 
         factor 1.5 in radius)
===> Real interval arrays automatically transformed to complex array 
         if a component is assigned a complex value (without warning)
===> Warning issued when complex interval defined by infsup(zinf,zsup)
===> Long arithmetic computations with error terms, i.e. valid long error bounds
===> Gradient derivative stored sparse for 50 and more unknowns
===> Warning issued when accessing .dx of sparse 2-dimensional gradient
===> Default sparse display of univariate polynomials
===> Default evaluation univariate polynomials by Horner's scheme
===> Default evaluation multivariate polynomials by sums of scaled powers
===> Polynomial function executed with warning if input polynomial does not depend on specified variable
===> Warning issued when accessing .s of sparse 2-dimensional slope
===> Hessian derivatives stored sparse for 10 and more unknowns
===> Warning issued when accessing .dx or .hx of sparse hessian array
 
=========================================================================
*****  Free for private and academic use. Commercial use or use in  *****
*****  conjunction with a commercial program which requires INTLAB  *****
*****  or part of INTLAB to functioning properly is prohibited.     *****
=========================================================================

Weiter kann mit help fortgefahren werden:

>> help

HELP topics:

toolbox/Intlab_V5.3  -  INTLAB  INTerval LABoratory. 
Intlab_V5.3/intval   - INTLAB interval toolbox
Intlab_V5.3/gradient - INTLAB interval gradient toolbox
Intlab_V5.3/hessian  - INTLAB interval hessian toolbox
Intlab_V5.3/slope    - INTLAB interval slope toolbox
Intlab_V5.3/polynom  - INTLAB polynom toolbox
Intlab_V5.3/utility  - INTLAB some utility routines
Intlab_V5.3/long     - INTLAB long toolbox (rudimentary, slow but correct)

Fragen zur Installation unter Linux beantwortet Ingo Tessmann, Tel.: 3187.