scikit-fuzzy Documentation

If you are re-implementing algorithms originally written in Matlab, be aware that scikit-fuzzy follows the IEEE standard for rounding.

IEEE rounding requires rounding to the nearest EVEN number when a value is exactly halfway between two integers. This is intended to minimize systematic rounding errors.

Examples:

• 1.5 $\rightarrow$ 2
• 2.5 $\rightarrow$ 2
• 3.5 $\rightarrow$ 4
• 4.5 $\rightarrow$ 4

Differences in results compared to Matlab are expected and are not considered bugs.

To generate the package logos, run the skfuzzy_logo.py script located in the directory. This produces two versions of the logo:

• logo.png: Low resolution, ideal for web use.
• logo_full.png: High resolution, ideal for print use.

python skfuzzy_logo.py

You can install the latest stable release of scikit-fuzzy from PyPI using pip. This command also upgrades existing installations to the latest version.

Dependencies:

• Matplotlib >= 3.1
• NumPy >= 1.6
• SciPy >= 0.9
• NetworkX >= 1.9

$ pip install -U scikit-fuzzy

If you want to contribute to the project or install from source, clone the repository and use one of the following commands:

• To install in editable mode: pip install -e .
• To install with development dependencies: pip install -e ".[develop]"

Alternatively, you can use the package without installing it by adding its path to your PYTHONPATH environment variable.

$ pip install -e . 
# or for development dependencies
$ pip install -e ".[develop]"

To generate the package icons (which contain only the 'S' and 'F' motifs), run the skfuzzy_icon.py script located in the directory. This produces three resolution versions for different platform needs:

• icon_32px.png: Low resolution, ideal for tab page icons.
• icon_64px.png: Medium resolution, good for small thumbnails or medium icons.
• icon_128px.png: High resolution, good for high PPI display thumbnails.

python skfuzzy_icon.py

In addition to the standard scikit-fuzzy package dependencies, generating these graphics requires the following Python packages:

• scikit-image >= 0.10
• matplotlib >= 1.2