[gmx-developers] Clusterizing a long-time trajectories

[David van der Spoel]

Den 2018-08-04 kl. 23:00, skrev Boris Timofeev:
> Dear David!
> 
> Option  -dr, it is new?
> Perhaps, you meant - dt? There is no wish to thin out a trajectory.
-dt indeed. The reason for long simulations is typically that 
transitions are slow and that it take long time to equilibrate. That 
also means that frames near each other are highly correlated and 
therefore the -dt option is no problem. However, this is obviously 
problem dependent.

It could be possible to devise some kind of incremental clustering 
algorithm that stores cluster centers and dynamically updates them. In 
that manner the whole trajectory does not need to be stored. However 
that would be a research project in itself, and I guess this kind of 
algorithms has been developed long time ago. If you find such an 
algorithm you could consider implementing it. Alternatively, if you see 
a way of getting around the N^2 RMSD matrix in the existing algorithms 
that would be interesting too.

> Sincerely,
> Boris.
> 
> 
>     Суббота, 4 августа 2018, 16:09 +03:00 от David van der Spoel
>     <spoel at xray.bmc.uu.se>:
> 
>     Den 2018-07-26 kl. 14:02, skrev Boris Timofeev:
>      > Dear developers!
>      >
>      > The existing realization of "gmx cluster" for calculations uses
>      > RMSD-matrix,
>      > quadratically increasing with increase in length of a trajectory,
>      > and already since 20000 frames calculation of clusters becomes
>     impossible
>      > because of requirements to memory and time.
>      >
>      > Whether there are some special methods for a clustering of
>     trajectories
>      > of big duration,
>      > containing 100000 and more frames?
>     Use the -dr flag.
>      >
>      > Best wishes
>      > Boris Timofeev.
>      >
>      >
>      >
>      >
>      >
> 
> 
>     -- 
>     David van der Spoel, Ph.D., Professor of Biology
>     Head of Department, Cell & Molecular Biology, Uppsala University.
>     Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
>     http://www.icm.uu.se
>     -- 
>     Gromacs Developers mailing list
> 
>     * Please search the archive at
>     http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
>     before posting!
> 
>     * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> 
>     * For (un)subscribe requests visit
>     https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>     or send a mail to gmx-developers-request at gromacs.org.
> 
> 
> 
> -- 
> Boris Timofeev
> 
> 


-- 
David van der Spoel, Ph.D., Professor of Biology
Head of Department, Cell & Molecular Biology, Uppsala University.
Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
http://www.icm.uu.se