mistify 0.0.1

Neuro-fuzzy systems with PyTorch.

Mistify

Mistify is a library built on PyTorch for building Neurofuzzy Systems. A neurofuzzy system is a trainable fuzzy system, typically consisting of a fuzzifier, rules layers, and a defuzzifier. Mistify provides a variety of modules to use at each layer of the pipeline

Installation

pip install mistify

Brief Overview

Mistify consists of subpackages for inference operations, fuzzification and defuzzification, and preprocessing and postprocessing. It incldes

• mistify: The core functions used for fuzzification and inference.
• mistify.fuzzify: Modules for building fuzzifiers and defuzzifiers. Has a variety of shapes or other fuzzification and defuzzification modules to use.
• mistify.infer: Modules for performing inference operations such as Or Neurons, Intersections, Activations etc.
• mistify.process: Modules for preprocessing or postprocessing on the data to input into the fuzzy system
• mistify.systems: Modules for building systems more easily.
• mistify.utils: Utilities used by other modules in Mistify.

Usage

Mistify's primary prupose is to build neurofuzzy systems or fuzzy neural networks using the the framework of PyTorch.

Here is a (non-working) example that uses alternating Or and And neurons.

class FuzzySystem(nn.Module):

    def __init__(
        self, in_features: int, h1: int, h2: int, out_features: int
    ):

        # Use for these builders for buliding a neuron
        # In this case, tehre is no wait fou
        AndNeruon = BuildAnd().no_wf().inter_on().prob_union()
        OrNeuron = BuildOr().no_wf().union_on().prob_inter()

        # 
        self.fuzzifier = mistify.fuzzify.SigmoidFuzzifier.from_linspace(
            n_terms, 'min_core', 'average'
        )
        self.flatten = FlattenCat()
        self.layer1 = OrNeuron(in_features * categories, h1)
        self.layer2 = AndNeruon(h1, h2)
        self.layer3 = OrNeuron(h2, out_features * out_categories)
        self.deflatten = DeflattenCat(out_categories)

        self.defuzzifier = mistify.fuzzify.IsoscelesFuzzyConverter.from_linspace(
            out_terms, 'min_core', 'average'
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:

        m = self.fuzzifier(x)
        m = self.flatten(m)
        m = self.layer1(m)
        m = self.layer2(m)
        m = self.layer3(m)
        # use to prepare for defuzzification
        m = self.deflatten(m)
        return self.defuzzifier.defuzzify(m)

Since it uses Torch, these fuzzy systems can easily be stacked.

Contributing

To contribute to the project

1. Fork the project
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a pull request

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

Citing this Software

If you use this software in your research, we request you cite it. We have provided a CITATION.cff file in the root of the repository. Here is an example of how you might use it in BibTeX: