CLUTO - Software for Clustering High-Dimensional Datasets

History of Changes

Ver: 2.1.2a, 1/9/07
  • Included build for Windows X86_64.
Ver: 2.1.2, 10/18/06
  • Eliminated the limits on the length of the line of the input files.
  • Fixed spelling errors in the -help for vcluster/scluster.
  • Eliminated the 32 bit limit on the size of the dynamically allocated memory. CLUTO can now take advantage of 64 bit address space machines.
  • Builds for OSX (powerpc and i386) and Linux x86_64.
  • Reduced memory requirements for post-clustering reordering of the cluster numbers.
  • Performance improvements for some hierarchical agglomerative schemes.
  • Experimental support for multi-core processors and SMPs using OpenMP for MS Windows and Linux-i686.
  • Redesigned the dynamic memory allocation scheme to be based on Doug Lea's malloc code.
  • An experimental set of new API calls is being provided that gracefully cleanup all internally allocated memory in case of critical errors and returns a code to the calling program indicating the type of the problem.
Ver: 2.1.1, 11/27/03
  • Fixed problems with the postscript output driver and newer versions of ghostscript.
  • Fixed problems with incorrect bitmap sizes for gif output format.
  • Fixed various font problems with the xfig output format.
  • Fixed a bug associated with incorrectly computing the similarities in the hierarchical tree files.
  • Corrected the ISIM values reported by the scluster program.
  • Fixed a bug in -showsummaries=itemsets.
  • Corrected the information displayed during -help for the -showsummaries option.
  • Reduced the memory requirements of the rb-based clustering methods.
Ver: 2.1, 8/26/02
  • Added an agglomerative algorithm that uses partitional-clustering to bias the agglomeration.
  • Added a -showsummaries option for analyzing the co-occurrence of the key features within the objects of each cluster.
  •  Added similarity function based on extended Jackard coefficient for graphs.
  • Added a cluster-selection method for repeated bisecting clustering that is based on the cardinality of the cluster's sub-space.
  • Fixed a major bug with Euclidean distance calculations on sparse datasets.
  • Streamlined some of the parameters of vcluster & scluster.
Ver: 2.0, 4/22/02
  • The new clustering programs are now called vcluster & scluster
  • Added graph-partitioning based clustering algorithms
  • Expanded all algorithms to operate on the object-to-object similarity graph.
  • Added -agglofrom and -agglocrfun functionality for hybrid clustering
  • Fixed numerous bugs :).
Ver: 1.5, 1/8/02
  • Added a set of fully functional agglomerative clustering algorithms.
  • Added support for hybrid partitional and agglomerative clustering.
  • Added an extensive set of cluster visualization capabilities.
  • Added support for correlation coefficient based similarities.
  • Added support for dense input files.