Python framework for writing programs to solve complex problems not amenable to brute force. The architecture is a descendant of Douglas Hofstadter and Melanie Mitchell's Copycat architecture.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

README.md

#PySeqsee A python framework for solving complex problems not amenable to brute force.

PySeqsee aims to be a robust framework for developing blackboard-architecture based programs that tackle hard problems in a human-like way.

It is open-source, under GNU GPLv3.

###Links

Mailing List:

Documentation: http://amahabal.github.com/PySeqsee/ Source Code: https://github.com/amahabal/PySeqsee Bug Tracker: https://github.com/amahabal/PySeqsee/issues Development Status: Alpha, but actively-developed working code

###Brief history and motivation

For over two decades now, Douglas Hofstadter's Fluid Analogies Research Group at Indiana University has designed computer simulations aimed at understanding human cognition. Each successive model has usually been written from scratch. Very little of the actual code from previous implementations was used by subsequent implementations, although ideas and the basic approach survived.

Not just were the implementations different, they were typically in different languages. Franz Lisp, Chez Scheme, C++, and even Perl have been used by various projects, and there was also talk of using Delphi. A Java port of Copycat exists.

This project aims to create a framework in which to implement various cognitive architectures. It is written in Python 3, and aims to provide many components out of the box without making too many irreversible commitments. That is, it provides a full suite of tools to get the job done, but also allows you to swap out any component and use the rest.

###Services provided (and their level of completion):

  • A reusable GUI. Every project will have a different workspace, but there is still much that is shared. PySeqsee allows you to just write the visualization for your data, and takes care of everything else. (complete)
  • A robust testing framework. If you wish to test how well a proposed new feature works, you can run a side-by-side comparison over many inputs and see the stats. All that is needed is a file with the inputs to test and in case the input is very specialized, a python class to convert these inputs to flags to be passed in. (status: functionality implemented, but statistical analysis to compare which of the two sides is better is not done).
  • A setup script that creates the skeleton of a new project. (Under development, basic functionality exists).
  • A coderack and facilities for writing codelets (complete).
  • A slipnet (known here as Long-term Memory), along with the ability to add nodes and edges and the ability to save it to disk (works, but could be significantly improved).
  • A stream of thought. This is a component that first appeared in Seqsee (written in Perl), but plays a central role here. The stream provides a temporal context --- that is, recent thoughts can influence what the system does in flexible ways. A full implementation exists.
  • A full reimplementation of Seqsee (status: under development. Many sequences seen, but not all that the perl version did). A short video of the Perl version can be found here: http://www.youtube.com/watch?v=2KWtRUg8kL8. The dissertation is here: http://www.amahabal.com/files/Seqsee--doublesided.pdf