Hi, Yes, it must be the same parameter vector. But if you have data at 2 field strengths, you have 2 R20a parameters in this model. Therefore the Jacobian requires two dR2eff/dR20a elements, but it currently has only one hard-coded. You need the self.end_index list in the target function class to correctly construct the Jacobian matrix. This is why it is very useful to have all the function, gradient, Hessian, and Jacobian functions/methods in this target function class. The covariance matrix function doesn't need to know this as you just pass in the NxM Jacobian and the MxM weighting matrix.
Regards, Edward On 2 September 2014 10:57, Troels Emtekær Linnet <[email protected]> wrote: > Hi Edward. > > For the Jacobian to work with minfx, the parameter vector send to the > function should match 1:1 with the one to > chi2 function. > > Best > Troels > > 2014-09-02 10:41 GMT+02:00 Edward d'Auvergne <[email protected]>: >> Hi Troels, >> >> I just noticed a problem here. For the Jacobian (as well as the >> gradient and Hessian), you will need one element of the dR2eff/dr20a >> gradient for each R20a parameter in the model. Same for the >> dR2eff/ddw gradient element. You currently return a list of 3 >> elements, but I'm guessing that the parameter vector sent into the >> target function is not simply [R20a, dw, k_AB]. >> >> Regards, >> >> Edward >> >> On 2 September 2014 10:29, <[email protected]> wrote: >>> Author: tlinnet >>> Date: Tue Sep 2 10:29:46 2014 >>> New Revision: 25529 >>> >>> URL: http://svn.gna.org/viewcvs/relax?rev=25529&view=rev >>> Log: >>> In the lib function of TSMFK01, returned the Jacobian as list of >>> derivatives. This is for easier manipulation in target function. >>> >>> task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR >>> estimation from Jacobian and Co-variance matrix of dispersion models. >>> >>> Modified: >>> branches/est_par_error/lib/dispersion/tsmfk01.py >>> >>> Modified: branches/est_par_error/lib/dispersion/tsmfk01.py >>> URL: >>> http://svn.gna.org/viewcvs/relax/branches/est_par_error/lib/dispersion/tsmfk01.py?rev=25529&r1=25528&r2=25529&view=diff >>> ============================================================================== >>> --- branches/est_par_error/lib/dispersion/tsmfk01.py (original) >>> +++ branches/est_par_error/lib/dispersion/tsmfk01.py Tue Sep 2 10:29:46 >>> 2014 >>> @@ -142,6 +142,8 @@ >>> @type k_AB: float >>> @keyword tcp: The tau_CPMG times (1 / 4.nu1). >>> @type tcp: numpy float array of rank [NE][NS][NM][NO][ND] >>> + @return: The Jacobian returned as list of derivatives. >>> This is for easier manipulation and possible back scaling from rad/s to >>> normal units in relax. >>> + @rtype: list of numpy arrays >>> """ >>> >>> # Get the partial derivatives. >>> @@ -149,7 +151,7 @@ >>> get_d_f_d_dw = d_f_d_dw(r20a=r20a, dw=dw, k_AB=k_AB, tcp=tcp) >>> get_d_f_d_k_AB = d_f_d_k_AB(r20a=r20a, dw=dw, k_AB=k_AB, tcp=tcp) >>> >>> - return transpose(array( [get_d_f_d_r20a , get_d_f_d_dw, >>> get_d_f_d_k_AB] ) ) >>> + return [get_d_f_d_r20a , get_d_f_d_dw, get_d_f_d_k_AB] >>> >>> >>> def d_f_d_r20a(r20a=None, dw=None, k_AB=None, tcp=None): >>> >>> >>> _______________________________________________ >>> relax (http://www.nmr-relax.com) >>> >>> This is the relax-commits mailing list >>> [email protected] >>> >>> To unsubscribe from this list, get a password >>> reminder, or change your subscription options, >>> visit the list information page at >>> https://mail.gna.org/listinfo/relax-commits >> >> _______________________________________________ >> relax (http://www.nmr-relax.com) >> >> This is the relax-devel mailing list >> [email protected] >> >> To unsubscribe from this list, get a password >> reminder, or change your subscription options, >> visit the list information page at >> https://mail.gna.org/listinfo/relax-devel _______________________________________________ relax (http://www.nmr-relax.com) This is the relax-devel mailing list [email protected] To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel
