Version 2.5

Minor change to previous version. Continuity seriation now includes sumgraph-by-weight output, which is not interesting to those who simply want seriations. It is interesting if you want to examine the details of how solutions get formed and pruned.

v2.4 Continuity and batch mode improvements

Minor release to improve batch mode operation (by including a batch builder that creates execution scripts usable on Grid Engine and other cluster managers). Also a minor cleanup of seriation statistics so that continuity and frequency seriation can operate independently (especially that you can do continuity seriation without doing frequency seriation).

v2.3 Bug fixes and performance enhancements

Xavi Rubio (xrubio) noted a bug in the published version and releases 2.1 and 2.2, in checking the ordering of triples. This has now been fixed. We did not directly integrate his pull request since he changed the way edge weights are handled and we had good reasons for the original method, and we may integrate his changes to the use of Multiprocessing for the seriationEvaluation workers in the next release after further testing. We also rewrote the spatial/geographic bootstrap significance test (enabled with --spatialsignificance and --spatialbootstrapN = 10000, etc) for much better performance (4.8x boost). We also cleaned up a bit of debugging output that wasn't switched off, and the program now cleans up its temporary pickle file directory, which is now located properly under /tmp. IDSS now tags every simulation run with a UUID, which is recorded along with the Git commit and release number in a metadata output file, AND into MongoDB database if you use the mongodb command line driver script instead of the plain driver.

IDSS V2.2 performance release

Version 2.2 of IDSS seriation preserves the same functionality as V2.1 (i.e., is focused mainly upon frequency seriation using the min-max graphs, but containing code for continuity and occurrence seriation). This version has a rewrite of the bootstrap confidence interval calculation to leverage NumPy and SciPy for better performance and future JIT compilation, and some adjustments to the parallel workers to minimize the duplication of unpickling of data. The resulting system is mostly bound by the efficiency of the NetworkX graph algorithms, and in particular the shortest path implementation.

IDSS Version 2.1

The core IDSS code is fully separated into a front-end script with command line processing, and backend libraries to facilitate further development. Some of the spatial bootstrap options are fixed and clarified since v2.0. This version was used to produce results for our SAA conference paper in April 2015.