[gmx-users] k-means clustering in GROMACS

[David van der Spoel]

Nikos Sgourakis wrote:
> Dear GROMACS users,
> 
> Here I have implemented the k-means clustering algorithm in the original 
> GROMACS module, gmx_cluster.c (called by the wrapper g_cluster).
> k-means is a heuristic algorithm that finds the partition of n points 
> (conformations) in k groups (clusters), such that the sum of the 
> distances of each point from the centroid of its belonging cluster is 
> minimized.
> To compile, re-run make after replacing the original file: `make g_cluster`
> Most of the in-line options of the original program work. However, there 
> are no output structure files. Instead, `./g_cluster -method kmeans etc` 
> will produce a file with the indeces of all conformations within each 
> one of the k clusters (k-means.dat), and a list of all centroids 
> (centroids.dat).  The conformations can then be easily retrieved from 
> the original trajectory file.
> Please, let me know of any difficulties in compiling/running kmeans for 
> GROMACS.
> 
> Nikos
> 
> 

Nikos,

first, thanks for contributing to gromacs. I have some questions about 
your code though. Have you run this code at all, and if so on what 
platform and with which compiler? The code contains this:

void kmeans(int k)
{
   int distr[k];

etc.

which is not standard C as far as I know. You can replace this by:

   int *dist;
   snew(distr,k);
   ....

   sfree(distr);

where snew and sfree are gromacs macros for allocating an array with 
length k for arbitrary  data types.

It would also be nice if you use a uniform indentation, preferable the 
emacs default.

Regards,
-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Assoc. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++