FuzzPy - Fuzzy Mathematics for Python

Overview

FuzzPy is a library for fuzzy sets, fuzzy graphs, and general fuzzy mathematics for Python.

Requirements

FuzzPy requires Python 2.6 or later (and supports Python 3.x).

For visualization (optional), one or more of the following are required:

• Gnuplot-py for fuzzy number visualization.
• PyDot for graph and fuzzy graph visualizations.

Epydoc is required for generating API documentation (optional).

Documentation and Examples

API documentation can be generated with Epydoc using python setup.py doc. This will create a subdirectory doc containing full API documentation in HTML format.

Examples demonstrating the use of FuzzPy can be found in the examples subdirectory.

Core Functionality

Unlike most other fuzzy libraries, FuzzPy focuses on pure fuzzy sets as its core functionality. The fset module provides a discrete fuzzy set class FuzzySet which behaves for the most part (and is a subclass of) the built-in Python set type.

The fgraph module provides the FuzzyGraph class, which is based on our own Graph class (in the graph module), and uses fuzzy sets for its vertex and edge sets. The graph modules also provide various combinatorial optimization and other graph-related algorithms.

The fnumber module provides a series of FuzzyNumber subclasses, for fuzzy subsets of the real numbers. The interface tries to mimic that of the discrete fuzzy set class as much as possible. This is intended to provide a basis for future work in fuzzy rule-based systems and the like.

Visualizations

Starting with v0.4.0, FuzzPy ships with a plugin-based visualization system, which can produce representations of most of the data structures supported by FuzzPy, and supports a variety of output formats.

One should consult the examples/visualizations.py to examine several different use cases, but the typical usage scenario works as such:

    from fuzz.visualization import VisManager

    # Create the appropriate plugin automatically
    vis = VisManager.create_backend(data_object)
    (vis_format, vis_data) = uvis.visualize()

    # Write data onto a file
    with open("output.%s" % vis_format, "wb") as fp:
        fp.write(vis_data)

Although the API is designed to be consistent across all plugins, we recommend consulting the docstrings and source code for plugins you decide to use in order to learn about the different way you can customize their behaviour, as well as their available output formats, supported data types, etc.

Development

FuzzPy is in a relatively early stage of development, and is changing rapidly. Functional compatibility may be broken between releases, even revisions in some cases. Once the project is reasonably feature-complete and best practices are established, the 1.x release series will begin, and the project will transition into a policy of not breaking functional compatibility across releases with the same minor version.

If you would like to help with development, please contact Aaron Mavrinac to get access to the GitHub repositories.

FuzzPy's main contributors are:

• Aaron Mavrinac mavrin1@uwindsor.ca
• Xavier Spriet linkadmin@gmail.com

Visualization Plugin Development

If none of the supplied data visualization plugins meet your specific need, it may be beneficial to write your own, and we encourage you to contribute your work back to the project if you do so.

Developing a new visualization plugin consists of the following steps:

• Create a new submodule for your plugin inside the vis_plugins directory
• import the AbstractPlugin abstract class from the abc_plugin submodule
  • from abc_plugin import AbstractPlugin
• Construct your plugin class. It must match the AbstractPlugin abstract class model.
• Add the following variables at the top of the submodule:
  • VIS_PLUGIN: Name of your plugin's class
  • VIS_TYPES: List of datatype classes supported
  • VIS_FORMATS: List of supported output formats for the plugin

Please consult abc_plugin.AbstractPlugin class as well as other example plugins such as the graph_pydot plugin for an explanation of the required methods and their behaviours.