README.md

MIPKmeans

A python implementation of MIP-Kmeans algorithm

Usage

usage: run.py [-h] [--ofile OFILE] [--sfile SFILE] [--n_rep N_REP]
              [--init {random,kmpp}] [--convergence {label,shift}]
              [--m_iter M_ITER] [--tol TOL] [--constraint_laziness {0,1,2,3}]
              dfile cfile k

Run MIP-Kmeans algorithm

positional arguments:
  dfile                 data file
  cfile                 constraint file
  k                     number of clusters

optional arguments:
  -h, --help            show this help message and exit
  --ofile OFILE         file to store the output
  --sfile SFILE         file to which to append the summary
  --n_rep N_REP         number of times to repeat the algorithm
  --init {random,kmpp}  initialization method
  --convergence {label,shift}
                        convergence criterion
  --m_iter M_ITER       maximum number of iterations of the main loop
  --tol TOL             tolerance for deciding on convergence
  --constraint_laziness {0,1,2,3}
                        whether to first add the constraints or only if they
                        are violated

To see a run of the algorithm on example data and constraints, run the script test.sh.

Dependencise

The program uses Gurobi solver to solve the assignment subproblem. You should also install the python interface to gurobi (the gurobipy module).