Dear Christoph I found a possible way to avoid the jumps in dihedral distributions. There are 3 points I should notice.
(1) The bin range of the histogram is made as [ x - 0.5 * step , x + 0.5 * step]. Therefore, when I choose min=-3.14, max=3.14, and step=0.04, the histogram gets sudden drops at both ends because there are no data in the ranges of [-3.16, -pi] and [pi, 3.16]. I think I should choose the values so that "min - 0.5 * step >= -pi" and "max + 0.5 * step <= pi". In addition, it is better to make (max - min)/step be integer. If not, the situation could become more complicated because the bin width of the histogram seems not to be the same as step value defined in the xml file. (2) The periodic option should be off. If there are data greater than max+0.5*step, they are counted in the periodic potion. For example, if I choose min=-3.12, min=3.12, and step=0.04 with the periodic option, the histogram again jumps at both data ends by the populations in [-pi, -3.14] and [3,14, pi]. The magnitude of the jumps is around 4% because (pi - 3.14)/0.04 ~ 0.0398. (3) There might be a possible bug in line 53 of tools/src/libtools/histogramnew.cc. I think floor function should be used instead of int as follows. i = (int)((v - _min)/ _step + 0.5); i = (int)floor((v - _min)/ _step + 0.5); Because i should be a negative integer if v is less than _min - 0.5*_step. But, for instance, int(-0.1) = 0 not -1. This could lead a jump at one end of the histogram because some data are over-counted as i=0. I found a similar problem in line 76 of tools/src/libtools/histogram.cc. Best regards Takahiro 2016年5月20日金曜日 17時08分09秒 UTC+2 Christoph Junghans: > > 2016-05-20 7:17 GMT-06:00 Takahiro Ohkuma <[email protected] > <javascript:>>: > > > > > > 2016年5月19日木曜日 19時20分44秒 UTC+2 Christoph Junghans: > >> > >> 2016-05-19 11:07 GMT-06:00 Takahiro Ohkuma <[email protected]>: > >> > Dear Christoph > >> > > >> > I am really grateful for writing a patch. > >> > Unfortunately, the patch seems not to help for my problem. > >> > I obtained almost the same results for the dihedral distribution and > the > >> > inverted potential. They still have jumps at the data ends. > >> > > >> > Just to make sure, I would like to write down the way I carried out > to > >> > install your patch. > >> > 1) I downloaded and install votca via "./build.sh --prefix=${prefix} > -ud > >> > tools csg" in a new directory. > >> Do 1.) and then > >> $ cd csg > >> $ git pull > >> $ git checkout origin/Dihedral > >> $ make > >> $ make install > >> You might have to do some debugging to find out where the jump comes > >> from, e.g. add a debug message when a point around -pi gets insert in > >> the histogram. > > > > Yes, I would try to do this. Could you tell me the place of codes in > which > > the histograms and potentials are computed during ibi? > > tools/src/libtools/histogram.cc? > Have a look here: > <https://github.com/votca/csg/blob/master/src/tools/csg_stat_imc.cc#L283> > > Christoph > > > > Takahiro > >> > >> > >> > 2) The two files of interaction.h and csg_stat_imc.cc files are > replaced > >> > with your patch. > >> > 3) I typed again "./build.sh --prefix=${prefix} -ud tools csg" to > enable > >> > the > >> > patch. > >> > 4) I sourced the new VOTCARC.bash before a simulation. > >> > > >> > Can I use a simplex method for bonded interactions such as > >> > Ryckaert-Bellemans functions? > >> It won't be hard, but you will have to do some coding! > >> You either need to write a file telling gnuplot how Ryckaert-Bellemans > >> are made up of cos functions or a parameter translator from VOTCA to > >> gromacs. > >> > >> Christoph > >> > > >> > > >> > Best regards > >> > Takahiro > >> > > >> > 2016年5月19日木曜日 0時01分51秒 UTC+2 Christoph Junghans: > >> >> > >> >> 2016-05-18 11:56 GMT-06:00 Takahiro Ohkuma <[email protected]>: > >> >> > Dear Christoph > >> >> > > >> >> > Thank you for your suggestions. Unfortunately, I found it was > >> >> > difficult > >> >> > for > >> >> > me to find out a clue in the votca source codes. > >> >> > > >> >> > In the csg_boltzmann conmmand, I could obtain a smooth profile on > >> >> > condition > >> >> > that "periodic = 1". There are no jumps at the data ends. > >> >> > Is it possible to turn on the periodic option during ibi process? > >> >> Can you check if the following patch helps: > >> >> > >> >> > >> >> < > https://github.com/votca/csg/commit/63f8d822187142dc82f3787dd2211a282e39b85b> > > >> >> > >> >> Christoph > >> >> > Or, can Votoca omit two data at both ends during ibi? > >> >> > > >> >> > Best regards > >> >> > Takahiro > >> >> > > >> >> > > >> >> > > >> >> > 2016年5月12日木曜日 19時07分37秒 UTC+2 Christoph Junghans: > >> >> >> > >> >> >> 2016-05-11 10:47 GMT-06:00 Takahiro Ohkuma <[email protected]>: > > >> >> >> > Dear Christoph > >> >> >> > > >> >> >> > Thank you very much for your prompt reply! > >> >> >> > I am doing short simulations becase I have not tried the > numbers > >> >> >> > you > >> >> >> > suggested. I would like to share the results once I get them. > >> >> >> > > >> >> >> > I have seen a similar behavior when I made the dihedral > potential > >> >> >> > through > >> >> >> > boltzmann inversion by using csg_boltzmann. > >> >> >> If this is happening in csg_boltzmann as well, it sounds like a > >> >> >> problem in the histogram class. > >> >> >> > >> >> >> > At that time, I clipped the edges as a poorly sampled region. > >> >> >> > > >> >> >> > Takahiro > >> >> >> > > >> >> >> > 2016年5月11日水曜日 17時17分56秒 UTC+2 Christoph Junghans: > >> >> >> >> > >> >> >> >> 2016-05-11 4:35 GMT-06:00 Takahiro Ohkuma <[email protected]>: > > >> >> >> >> > Dear all > >> >> >> >> > > >> >> >> >> > I am trying IBI for the bonded interactions of my polymer > >> >> >> >> > because > >> >> >> >> > the > >> >> >> >> > bonded > >> >> >> >> > potentials obtained by a single-chain simulation in vacuum > did > >> >> >> >> > not > >> >> >> >> > well > >> >> >> >> > reproduce the bond distributions in the melt. > >> >> >> >> > > >> >> >> >> > According to the tutorial of "hexane/ibi_bonded", I made a > >> >> >> >> > setting.xml > >> >> >> >> > and > >> >> >> >> > prepared other files. > >> >> >> >> > But, at the calculation of dihedral distribution in the IBI > >> >> >> >> > steps, > >> >> >> >> > I > >> >> >> >> > always > >> >> >> >> > meet a strange sharp peak at an edge of the distribution > (one > >> >> >> >> > data > >> >> >> >> > point) as > >> >> >> >> > the attached png file. This gives a sharp dihedral potential > at > >> >> >> >> > the > >> >> >> >> > edge, > >> >> >> >> > and then the next simulation often diverges. > >> >> >> >> > Although I tried other values of <min>, <max>, and <step>, > the > >> >> >> >> > situation > >> >> >> >> > did > >> >> >> >> > not changed. > >> >> >> >> > Is there any ways to manage this peak in the setting.xml? > >> >> >> >> Your settings.xml look correct to me! > >> >> >> >> > >> >> >> >> IBI for dihedrals isn't very well test and it looks to me > there > >> >> >> >> is > >> >> >> >> an > >> >> >> >> issue with boundary conditions, i.e. dihedrals are periodic. > >> >> >> >> Did you try to set max=-min=2.5, or min=0.5 and max= 2pi-0.5? > >> >> >> >> > >> >> >> >> Looking at the distribution, it seems the bug is in csg_stat > >> >> >> >> itself. > >> >> >> >> I > >> >> >> >> guess the problem is here: > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> < > https://github.com/votca/csg/blob/master/src/tools/csg_stat_imc.cc#L400> > >> >> >> >> It is marked as a TODO ;-) > >> >> >> >> However, the normalization should not matter as IBI does a > ratio > >> >> >> >> of > >> >> >> >> the distributions, but for the periodic histograms there might > be > >> >> >> >> a > >> >> >> >> round off error. > >> >> >> >> > >> >> >> >> Christoph > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> > > >> >> >> >> > Best regards > >> >> >> >> > Takahiro > >> >> >> >> > > >> >> >> >> > -- > >> >> >> >> > 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 https://groups.google.com/group/votca. > >> >> >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> -- > >> >> >> >> 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 https://groups.google.com/group/votca. > >> >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> >> > >> >> >> > >> >> >> > >> >> >> -- > >> >> >> 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 https://groups.google.com/group/votca. > >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> > >> >> > >> >> > >> >> -- > >> >> 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 https://groups.google.com/group/votca. > >> > For more options, visit https://groups.google.com/d/optout. > >> > >> > >> > >> -- > >> Christoph Junghans > >> Web: http://www.compphys.de > > > > > > 2016年5月19日木曜日 19時20分44秒 UTC+2 Christoph Junghans: > >> > >> 2016-05-19 11:07 GMT-06:00 Takahiro Ohkuma <[email protected]>: > >> > Dear Christoph > >> > > >> > I am really grateful for writing a patch. > >> > Unfortunately, the patch seems not to help for my problem. > >> > I obtained almost the same results for the dihedral distribution and > the > >> > inverted potential. They still have jumps at the data ends. > >> > > >> > Just to make sure, I would like to write down the way I carried out > to > >> > install your patch. > >> > 1) I downloaded and install votca via "./build.sh --prefix=${prefix} > -ud > >> > tools csg" in a new directory. > >> Do 1.) and then > >> $ cd csg > >> $ git pull > >> $ git checkout origin/Dihedral > >> $ make > >> $ make install > >> You might have to do some debugging to find out where the jump comes > >> from, e.g. add a debug message when a point around -pi gets insert in > >> the histogram. > >> > >> > 2) The two files of interaction.h and csg_stat_imc.cc files are > replaced > >> > with your patch. > >> > 3) I typed again "./build.sh --prefix=${prefix} -ud tools csg" to > enable > >> > the > >> > patch. > >> > 4) I sourced the new VOTCARC.bash before a simulation. > >> > > >> > Can I use a simplex method for bonded interactions such as > >> > Ryckaert-Bellemans functions? > >> It won't be hard, but you will have to do some coding! > >> You either need to write a file telling gnuplot how Ryckaert-Bellemans > >> are made up of cos functions or a parameter translator from VOTCA to > >> gromacs. > >> > >> Christoph > >> > > >> > > >> > Best regards > >> > Takahiro > >> > > >> > 2016年5月19日木曜日 0時01分51秒 UTC+2 Christoph Junghans: > >> >> > >> >> 2016-05-18 11:56 GMT-06:00 Takahiro Ohkuma <[email protected]>: > >> >> > Dear Christoph > >> >> > > >> >> > Thank you for your suggestions. Unfortunately, I found it was > >> >> > difficult > >> >> > for > >> >> > me to find out a clue in the votca source codes. > >> >> > > >> >> > In the csg_boltzmann conmmand, I could obtain a smooth profile on > >> >> > condition > >> >> > that "periodic = 1". There are no jumps at the data ends. > >> >> > Is it possible to turn on the periodic option during ibi process? > >> >> Can you check if the following patch helps: > >> >> > >> >> > >> >> < > https://github.com/votca/csg/commit/63f8d822187142dc82f3787dd2211a282e39b85b> > > >> >> > >> >> Christoph > >> >> > Or, can Votoca omit two data at both ends during ibi? > >> >> > > >> >> > Best regards > >> >> > Takahiro > >> >> > > >> >> > > >> >> > > >> >> > 2016年5月12日木曜日 19時07分37秒 UTC+2 Christoph Junghans: > >> >> >> > >> >> >> 2016-05-11 10:47 GMT-06:00 Takahiro Ohkuma <[email protected]>: > > >> >> >> > Dear Christoph > >> >> >> > > >> >> >> > Thank you very much for your prompt reply! > >> >> >> > I am doing short simulations becase I have not tried the > numbers > >> >> >> > you > >> >> >> > suggested. I would like to share the results once I get them. > >> >> >> > > >> >> >> > I have seen a similar behavior when I made the dihedral > potential > >> >> >> > through > >> >> >> > boltzmann inversion by using csg_boltzmann. > >> >> >> If this is happening in csg_boltzmann as well, it sounds like a > >> >> >> problem in the histogram class. > >> >> >> > >> >> >> > At that time, I clipped the edges as a poorly sampled region. > >> >> >> > > >> >> >> > Takahiro > >> >> >> > > >> >> >> > 2016年5月11日水曜日 17時17分56秒 UTC+2 Christoph Junghans: > >> >> >> >> > >> >> >> >> 2016-05-11 4:35 GMT-06:00 Takahiro Ohkuma <[email protected]>: > > >> >> >> >> > Dear all > >> >> >> >> > > >> >> >> >> > I am trying IBI for the bonded interactions of my polymer > >> >> >> >> > because > >> >> >> >> > the > >> >> >> >> > bonded > >> >> >> >> > potentials obtained by a single-chain simulation in vacuum > did > >> >> >> >> > not > >> >> >> >> > well > >> >> >> >> > reproduce the bond distributions in the melt. > >> >> >> >> > > >> >> >> >> > According to the tutorial of "hexane/ibi_bonded", I made a > >> >> >> >> > setting.xml > >> >> >> >> > and > >> >> >> >> > prepared other files. > >> >> >> >> > But, at the calculation of dihedral distribution in the IBI > >> >> >> >> > steps, > >> >> >> >> > I > >> >> >> >> > always > >> >> >> >> > meet a strange sharp peak at an edge of the distribution > (one > >> >> >> >> > data > >> >> >> >> > point) as > >> >> >> >> > the attached png file. This gives a sharp dihedral potential > at > >> >> >> >> > the > >> >> >> >> > edge, > >> >> >> >> > and then the next simulation often diverges. > >> >> >> >> > Although I tried other values of <min>, <max>, and <step>, > the > >> >> >> >> > situation > >> >> >> >> > did > >> >> >> >> > not changed. > >> >> >> >> > Is there any ways to manage this peak in the setting.xml? > >> >> >> >> Your settings.xml look correct to me! > >> >> >> >> > >> >> >> >> IBI for dihedrals isn't very well test and it looks to me > there > >> >> >> >> is > >> >> >> >> an > >> >> >> >> issue with boundary conditions, i.e. dihedrals are periodic. > >> >> >> >> Did you try to set max=-min=2.5, or min=0.5 and max= 2pi-0.5? > >> >> >> >> > >> >> >> >> Looking at the distribution, it seems the bug is in csg_stat > >> >> >> >> itself. > >> >> >> >> I > >> >> >> >> guess the problem is here: > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> < > https://github.com/votca/csg/blob/master/src/tools/csg_stat_imc.cc#L400> > >> >> >> >> It is marked as a TODO ;-) > >> >> >> >> However, the normalization should not matter as IBI does a > ratio > >> >> >> >> of > >> >> >> >> the distributions, but for the periodic histograms there might > be > >> >> >> >> a > >> >> >> >> round off error. > >> >> >> >> > >> >> >> >> Christoph > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> > > >> >> >> >> > Best regards > >> >> >> >> > Takahiro > >> >> >> >> > > >> >> >> >> > -- > >> >> >> >> > 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 https://groups.google.com/group/votca. > >> >> >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> >> >> > >> >> >> >> > >> >> >> >> > >> >> >> >> -- > >> >> >> >> 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 https://groups.google.com/group/votca. > >> >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> >> > >> >> >> > >> >> >> > >> >> >> -- > >> >> >> 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 https://groups.google.com/group/votca. > >> >> > For more options, visit https://groups.google.com/d/optout. > >> >> > >> >> > >> >> > >> >> -- > >> >> 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 https://groups.google.com/group/votca. > >> > For more options, visit https://groups.google.com/d/optout. > >> > >> > >> > >> -- > >> 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] <javascript:>. > > To post to this group, send email to [email protected] > <javascript:>. > > Visit this group at https://groups.google.com/group/votca. > > For more options, visit https://groups.google.com/d/optout. > > > > -- > 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 https://groups.google.com/group/votca. For more options, visit https://groups.google.com/d/optout.
