First Stable Release

XCS Project Home

Changes:

• Added unit testing.
• Added type checking.
• Significant speed improvements.
• Tutorial update.
• Major changes to interfaces and class/package names. (You'll want to read the updated tutorial.)
• Vastly improved in-source documentation.
• Various bugfixes.

The package has been made available as a wheel on pypi under the Revised BSD License and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

XCS Project Home

Changes:

• Updated README to point to the HTML version of the tutorial on pythonhosted.org
• Interface change - situation tracking
  • Added situation argument to ActionSet.__init__()
  • Added situation property to ActionSet
  • Removed now unnecessary situation argument to LCSAlgorithm.update()
• Added BitString.random() to allow numpy to be used in bitwise mutations
  when it is available.
• Added BitString.crossover_template() to allow numpy to be used in bit
  condition crossover when available.
• Added initialization from ordinary strings to __init__ methods of BitString and BitCondition
• Added the ability to detect and switch between numpy versus pure Python BitString
  implementations.
• Fixed a bug in BitString.__getitem__ in both implementations which was
  causing the underlying array type to leak through when using slices.
• Added the abstract reset() method to OnLineProblem, and implemented it
  for each problem class.
• Added the HaystackProblem, which requires the algorithm to search for a
  single relevant bit among a large set of irrelevant ones.
• Fixed a bug where the wildcard probability was being inverted in the
  BitCondition.cover() method.
• Renamed the tutorial to XCSTutorial
• Added new section to the tutorial, Defining New Problem Types

The package has been made available as a wheel on pypi under the Revised BSD License and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

XCS Project Home

Changes:

• Fixed file encoding issue which was preventing installation for some users.

The package has been made available as a wheel on pypi under the Revised BSD License and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

XCS Project Home

Changes:

• Improved README file
• Improved tutorial
• Added tutorial to distribution files
• Updated project home

The package has been made available as a wheel on pypi under the Revised BSD License and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

Changes:

• Major bug fix to resolve population size issue
• Added tutorial
• Documentation updates
• Significant changes to interface to separate algorithm from support functionality
• Replaced prints with logging

The package has been made available as a wheel on pypi and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

Enhancements continue in preparation for initial release of version 1.0.0.

Changes:

• Refactored to separate problems and problem interface from the algorithm itself
• Improved documentation
• Added iteration over Populations
• Added pretty-printing of Populations
• Improved version tracking mechanism

The package has been made available as a wheel on pypi and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

The canonical XCS algorithm has now been fully implemented, including both GA and action set subsumption. Basic testing has been performed to ensure that the code works for Anaconda 3.3 and Anaconda 3.4 on Windows 8, both with and without numpy 1.9.2 installed.

The package has been made available as a wheel on pypi and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.

Initial pre-release of the XCS (Accuracy-based Classifier Systems) library. The code is mostly untested and some functionality is only partially implemented. However, it does run successfully in an end-to-end testing scenario on the author's system (Anaconda 3.4 on Windows 8) and should be immediately usable to anyone wanting to explore the XCS algorithm.

The package has been made available as a wheel on pypi and can be installed with pip using the command pip install xcs. It can also be downloaded directly from the python package index.