Hi Troels, I'm not sure about the kex >= 1e5 check. You have to be careful, because a number of the analytic models - CR72 possibly included - are not defined for such timescales (of course this is dependant on the alpha scale, hence dw has an effect). So there are a series of models where kex >= 1e5 does not return R20!
Note that if you force the return of R20 when kex >= 1e5 when this is not valid, then you will introduce a discontinuity into the chi-squared space (https://en.wikipedia.org/wiki/Discontinuity_%28mathematics%29). Discontinuities are 100% fatal for all the local optimisers in minfx (https://gna.org/projects/minfx/). Therefore it is best to avoid a check of kex >= 1e5 to force a return of R20. Cheers, Edward On 26 May 2014 13:38, <[email protected]> wrote: > Author: tlinnet > Date: Mon May 26 13:38:09 2014 > New Revision: 23414 > > URL: http://svn.gna.org/viewcvs/relax?rev=23414&view=rev > Log: > Critical fix for the math domain catching of model CR72. > > This was discovered with the added 8 unit tests demonstrating edge case 'no > Rex' failures. > > This follows from the ideas in the post > http://article.gmane.org/gmane.science.nmr.relax.devel/5858. > This is related to: task #7793: (https://gna.org/task/?7793) Speed-up of > dispersion models. > > This is to implement catching of math domain errors, before they occur. > > When kex is large, ex: kex = 1e5, the values of: > etapos = eta_scale * sqrt(Psi + sqrt_psi2_zeta2) / cpmg_frqs > > will exceed possible numerical representation. > > The cathing of these occurences needed to be re-written. > > Modified: > branches/disp_speed/lib/dispersion/cr72.py > > Modified: branches/disp_speed/lib/dispersion/cr72.py > URL: > http://svn.gna.org/viewcvs/relax/branches/disp_speed/lib/dispersion/cr72.py?rev=23414&r1=23413&r2=23414&view=diff > ============================================================================== > --- branches/disp_speed/lib/dispersion/cr72.py (original) > +++ branches/disp_speed/lib/dispersion/cr72.py Mon May 26 13:38:09 2014 > @@ -128,9 +128,9 @@ > k_BA = pA * kex > k_AB = pB * kex > > - # Catch parameter values that will result in no exchange, returning flat > R2eff = R20 lines. > - if dw == 0.0 or pA == 1.0 or kex == 0.0: > - return array([r20_kex]*num_points) > + # Catch parameter values that will result in no exchange, returning flat > R2eff = R20 lines (when kex = 0.0, k_AB = 0.0). > + if dw == 0.0 or pA == 1.0 or k_AB == 0.0 or kex >= 1.e5: > + return array([r20a]*num_points) > > # The Psi and zeta values. > if r20a != r20b: > @@ -156,16 +156,12 @@ > # Catch math domain error of cosh(val > 710). > # This is when etapos > 710. > if max(etapos) > 700: > - R2eff = array([1e100]*num_points) > - > - return R2eff > + return array([r20a]*num_points) > > # The arccosh argument - catch invalid values. > fact = Dpos * cosh(etapos) - Dneg * cos(etaneg) > if min(fact) < 1.0: > - R2eff = array([r20_kex]*num_points) > - > - return R2eff > + return array([r20_kex]*num_points) > > # Calculate R2eff. > R2eff = r20_kex - cpmg_frqs * arccosh( fact ) > > > _______________________________________________ > 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
