GASTON - Download

Mining Structured Data
You can download the algorithm with occurrence lists here:


A version without occurrence lists is available here:


Changes since gaston-1.0.tar.gz:
  • bugfix: some very cyclic frequent graphs were not found or found too often with some datasets; these issues have been solved, as far as we are aware of them
Changes since gaston-0.1.tar.gz:
  • additional tests added to avoid the computation of occurrence list joins and extensions. Slightly increased performance due to do this change.
  • code cleanup.
Changes since gaston-0.2.tar.gz:
  • bugfix: print correct node labels
  • bugfix: some frequent graphs were not found due to an implementation glitch, this is corrected now
  • additional feature: stop searching in one of the phases (to find only frequent paths or free trees)
Changes since gaston-0.3.tar.gz:
  • optimisation: the dependency on Nauty has been removed. Gaston now uses its own canonical form which in practice is much more efficient. PTE runtimes have been lowered by 1/3.
  • bugfix: some frequent cyclic graphs were not found due to a bug in the computation of occurrence list joins. The problem did not surface in most tests, but is corrected now. Performance is in practice not affected.
  • additional feature: a version without occurrence lists is now also available. It recomputes occurrences with a brute-force search. Runtimes are higher, but memory usage is considerably lower. The NCI dataset (250.251 examples) is easily minable within 160MB of main memory.
  • additional feature: one can now define a maximum size on the frequent graphs.

E-mail:snijssen at