Is an implementation of K-means algorithm in Python.

There are two files, the paralel code and the serial code, most of the Prerequisites and instalation steps are just applicable to the paralel code.

Install openMPI (Ubuntu):

$ sudo apt-install openmpi-bin openmpi-doc libopenmpi-dev

• Anaconda (Python3) -> 5.0.1
• Python -> 2.7.13
• openMPI -> 1.6.3

For python you need something like Anaconda, miniconda or virtualenv for installing the packages just for the project.

$ conda create -n mpi python=2.7
$ conda install -n mpi numpy
$ conda install -n mpi mpi4pi

Then

$ source activate mpi
(mpi) $ mpiexec --version
> mpiexec (OpenRTE) 1.6.3

At the end of the definitions of the methods you should adapt this 2 variables:

k = <Number of clusters>
datasetLocation = <location of the dataset>

(mpi) $ mpiexec -n <number of cores> python ./Paralel.py #Paralel

$ python ./Serial #Serial

> Clusters: <Array of clusters>
Documents with no Relation: <Set of documents of no relation>
Time: <Time in seconds>

• Diego Alejandro Perez
• Edwin Montoya Jaramillo

The Department of computer science of EAFIT University has a machine with multiple cores where we can test the power of Paralel as shown below:

$ ssh <VPN Username>@192.168.10.115
>password:*********
<VPN Username>@hpcdis:~/$<Download Repo>
<VPN Username>@hpcdis:~/$cd k-meansHPC/
<VPN Username>@hpcdis:~/$source activate mpi
(mpi) <VPN Username>@hpcdis:~/$ mpiexec -n <# of cores you want. Max 70> python ./Paralel.py

• Edwin Nelson Montoya Múnera
• Juan David Pineda Cardenas
• Juan Francisco Cardona Mc'Cormick
• Daniel Hoyos Ospina
• Daniela Serna Escobar
• Daniel Rendon