KNN Algorithm with MPI

Implements kNN algorith both sequentially and with MPI for distributed systems.

Dependencies

• MPI library and mpic++ wrapper of compiler gcc. Can be installed by running:

  sudo apt-get install libopenmpi-dev
  

• gsl/gsl_cblas library: needed for matrix calculations. Can be installed by running:

  sudo apt-get install libgsl-dev
  

Running

• Clone repo and run make
• fileProducer is used to produce random data sets to test the programm. Usage:

  ./fileProducer <outputfilename> <numberOfPoints> <numberOfDimensions>
  

  e.g.

  ./fileProducer data.txt 1000 100
  

  Produces a file with 1000 random points with 100 dimensions. The output file format (and the expected from the programm) is:

  <numberOfPoints> <numberOfDimensions>
  d1 d2 ... dn
  d1 d2 ... dn 
  .
  .
  .
  d1 d2 ... dn
  

• mainSeq is the sequential implementation of the algorithm and can be run like this:

  ./mainSeq <dataFile> <numberOfNeighbors>
  

  e.g.

  ./mainSeq data.txt 15
  

• mainMPI is the distributed implementation of the algorithm and can be run like this:

  mpirun -np <numberOfProcesses> ./mainMPI <dataFile> <numberOfNeighbors>
  

  or

  mpirun --use-hwthread-cpus ./mainMPI <dataFile> <numberOfNeighbors>