IPP Software Navigation Tools IPP Links Communication Pan-STARRS Links

Ignore:
Timestamp:
Jul 19, 2007, 2:22:24 PM (19 years ago)
Author:
magnier
Message:

dLinear, the linearized change in the chisq, was being calculated
AFTER the parameters were set to their limits. this is inconsistent
with the assumptions, and resulted in impossible negative values for
dLinear, which in turn caused problems for the fitting convergence.

I moved the call to dLinear into GuessABP, and modified the APIs a bit
to deal with this.

File:
1 edited

Legend:

Unmodified
Added
Removed
  • trunk/psLib/src/math/psMinimizeLMM.h

    r13981 r14320  
    77 * @author GLG, MHPCC
    88 *
    9  * @version $Revision: 1.10 $ $Name: not supported by cvs2svn $
    10  * @date $Date: 2007-06-26 19:27:04 $
     9 * @version $Revision: 1.11 $ $Name: not supported by cvs2svn $
     10 * @date $Date: 2007-07-20 00:22:24 $
    1111 * Copyright 2004-2005 Maui High Performance Computing Center, University of Hawaii
    1212 */
     
    165165    const psVector *paramMask,
    166166    psMinimizeLMLimitFunc checkLimits,
    167     psF32 lambda
     167    psF32 lambda,
     168    psF32 *dLinear
    168169);
    169170
     
    171172    const psVector *Beta,
    172173    const psVector *beta,
     174    const psVector *paramMask,
    173175    psF32 lambda);
    174176
Note: See TracChangeset for help on using the changeset viewer.