2013/3/28 massimo sandal <[email protected]>: > Hi, > > We're having some doubts on how post-processing should be handled during > iterative procedures to get potentials. From the manual of VOTCA > ("Post-processing of the potential"), we find three procedures: > > Clipping: "Regions with an irregular distribution of samples should be > deleted first [...] Alternatively, manually check the range where the > potential still looks good and is not too noisy and set the flags in the > potential file of the bad parts by hand to o (out of range). Those values > will later be extrapolated and overwritten" > > Resampling: "Use the command > csg_resample --in table.pot --out table_resample.pot \ --grid min:step:max > > [...] It is important to note that the values min and max don’t correspond > to the minimum and maximum value in the input file, but to the range of > values the potential is desired to cover after extrapolation. Therefore, > values in [min,max] that are not covered in the file are automatically > marked by a flag o (for out of range) for extrapolation in the next step." > > Extrapolation: "The following line > csg_call table extrapolate [options] table_resample.pot \ > table_extrapolate.pot > > calls the extrapolation procedure, which processes the range of values > marked by csg_resample. The input file is table_resample.pot created in the > last step. > > After resampling, all values in the potential file that should be used as a > basis for extrapolation are marked with an i, while all values that need > extrapolation are marked by o. [...] The output table_extrapolate.pot of the > extrapolation step can now be used for the coarse-grained run. [...]" > > > Now, all of this is described as something that you do manually after you > had your final "raw" potential from a procedure like Boltzmann inversion. > However, within an iterative procedure like IBI, I would expect this to be > done for every iteration as well, so it has to be somehow automated in the > procedure -if artefacts or jumps due to poorly sampled regions etc. > accumulated, I fear that something could go wrong. And in fact the VOTCA > manual says "These scripts are the same ones as those used for iterative > methods in chapter 7" You should choose min and max of the interactions such that they are sampled sufficiently, otherwise the update for certain values will depend on poor statistic, which makes the results unpredictable. > > We however don't understand if and how it happens. Looking at 1.3 tutorial > files included in the VOTCA distributions, for the coarse-graining of water > with IBI (spce/ibi) we found this in the settings.xml : > > <inverse> > <!-- target distribution (rdf), just give gromas rdf.xvg --> > <target>CG-CG.dist.tgt</target> > <!-- update cycles --> > <do_potential>1</do_potential> > <!-- additional post processing of dU before added to potential --> > <post_update></post_update> > <!-- additional post processing of U after dU added to potential --> > <post_add></post_add> > > > but the <post_update> and <post_add> tags are empty. We have no idea > therefore: > - if we have to put something there > - if yes, what and where (what is to be done post_update but not post_add, > for example?) Is there some settings.xml with examples of this? What is > usually needed? Section 7.4.3 has some examples. (Also see the archive of the mailing list.)
In short: post-update scripts act on delta_U while post_add scripts act on U_new in every iteration. post-update scripts have "postupd" as key1 (see table in section 10.4), while postadd scripts have "postadd" as key1. Useful post-update scripts are scaling, smoothing or pressure correction of delta_U Useful postadd scripts are for calculating convergence or plotting of U_n+1. Christoph > > Can you give some hint here? Thanks a lot! > > Dr. Massimo Sandal > > -- > You received this message because you are subscribed to the Google Groups > "votca" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/votca?hl=en. > For more options, visit https://groups.google.com/groups/opt_out. > > -- Christoph Junghans Web: http://www.compphys.de -- You received this message because you are subscribed to the Google Groups "votca" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/votca?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
