README

This project implements canopy clustering as described by [1]. The
clustering is performed over a list of string sequences using a
compression-based distance function [2]. The code is implemented in
C++ using boost libraries and OpenMP for the parallelisation of the
distance computations.


CONFIGURATION

OpenMP can be enabled using the --enable-openmp flag at the
configuration step.


EXECUTION

The following command-line parameters are provided:

General Configuration:
  --help                produce help message
  --version             show the version

I/O Configuration:
  --result arg (=./results) results directory.
  --sequence arg            sequence file.

Canopy Configuration:
  --t1 arg (=0.25)      t2
  --t2 arg (=0.5)       t1, where t1 > t2
  --sample arg (=0)     Sample distance calculations
  --pairs arg           pairs for distance calculations

The sequence file has two columns separated by a comma. The first one
is the identifier for the sequence and the second column is the
string sequence.

If a pairs file is specified given the format i,j where i and j are
sequence identifiers, then the distance for those pairs is
computed. If a sample integer N is provided then the distances between
N sequences is computed. Otherwise, the canopies are computed over all
sequences and an output is produced in the format i,j,d where i and j
are sequence identifiers and d is the distance between those. Only
one-directional distances are computed (upper triangular form) even
though the NCD distance computation [2] may not be exactly
equal. However, for large enough sequences this in-equality is
negligible.



[1] Efficient clustering of high-dimensional data sets with
    application to reference matching
    by: Andrew McCallum, Kamal Nigam, Lyle H. Ungar
    In Proceedings of the sixth ACM SIGKDD international conference on
    Knowledge discovery and data mining (2000), pp. 169-178,
    doi:10.1145/347090.347123

[2] Clustering by compression
    by: R. Cilibrasi, P. M. B. Vitanyi
    Information Theory, IEEE Transactions on, Vol. 51, No. 4. (April
    2005), pp. 1523-1545, doi:10.1109/tit.2005.844059