Although the first commit to dit was in January of 2013, a much longer
history has influenced the design of dit. Here we provide that historical
context.
Originally, dit began as a single module in cmpy, a
Python package for computational mechanics developed as part of the
Complexity Sciences Center at UC Davis. In a sense, this was version -2.0.
The commit which introduced information.py (on 2007-04-18) included the
following statement:
Introducing some functions that compute information-theoretic quantities. In the future, we might need to formalize a 'distribution' into a dictionary-like class.
It was intentionally simple and cludgy, but it did what it needed to
do---providing functions for calculating entropy, mutual information,
conditional entropy, etc. As cmpy grew, this single module eventually
became a subpackage which included even more functionality. However, it was
already abundantly clear that a more formal information theory package
would be needed.
This brings us to a fairly innocent commit message in January of 2010:
first stab at a Distribution class
This was a complete rewrite and, in hindsight, could be seen as dit -1.0.
It was fresh, formal and exactly what cmpy needed. Its development
continued and was eventually split off into a new infotheory subpackage. With
the subpackage in place, features blossomed over the next few years to include
multivariate generalizations of Shannon's information theory as well algorithms
for calculating candidates measures for redundancy, synergy, and unique
information.
Still, the growth of the infotheory subpackage was quite organic, and many
of the original designs has been diluted due to unexpected use cases.
Discussions on a rewrite were commonplace, but design is always tricky and so,
a rewrite was delayed for some time.
In September of 2012, the first bits of dit were discussed over GChat
(now known as Google Hangouts), and written to file. Then finally, the
first commit made its way into the record books in January of 2013, now
available as a completely separate package. Its development was slow but
steady, and picked up quite a bit around September 2013.
This short outline is no where near detailed enough to give justice to the
work and decisions made along the way, but dit has benefited immensely from
these prior experiences---they have helped determine the API, typical use
patterns, and much more. Given the history, its not unlikely that dit might
eventually be rewritten as well. Until then, it will grow organically
to meet the needs of those who use it.
dit and those who have contributed to dit have received support throughout
the years from a variety of sources. We list them below. If you have
contributed to dit and a support acknowledgment does not appear below, please
help us remedy the situation, and similarly, please let us know if you'd like
something modified or corrected.
Complexity Sciences CenterDepartment of PhysicsUniversity of California-DavisPI: James P. CrutchfieldCenter for Complexity and Collective ComputationWisconsin Institute for DiscoveryUniversity of Wisconsin-MadisonPI: David C. Krakauer and Jessica C. FlackDepartment of Computer ScienceUniversity of Colorado-BoulderPI: Elizabeth Bradley
ARO grant W911NF-12-1-0288DARPA Physical Intelligence Subcontract No. 9060-000709NSF Grant No. PHY-0748828John Templeton Foundation