A Python implementation of the Causal State Splitting Reconstruction (CSSR) algorithm for inferring epsilon-transducers from data generated by discrete-valued, discrete-time input/output systems. This page provides the full source code, released under the GNU Public License.
This implementation has been tested on Mac OS X 10.10.3, but should work on any Unix-based OS. It requires the following Python dependencies:
These packages may be installed individually. A useful collection of these and related libraries for scientific computing with Python may be installed using Enthought or Anaconda.
Visualization of the epsilon-machines requires Graphviz or similar software for reading dot files.
NOTE: This is a legacy version of transCSSR for use with Python 2.7. The most recent version of transCSSR for use with Python 3.7+ is hosted here.