highfis 0.3.0

highFIS is a comprehensive Python package for training and evaluating high-dimensional TSK fuzzy systems, built on PyTorch and compatible with the scikit-learn API.

highFIS

Python library for high-dimensional Takagi–Sugeno–Kang (TSK) fuzzy inference systems, built on PyTorch with a scikit-learn compatible API.

📦 Installation

Install from PyPI:

pip install highfis

🧠 Quick Start

from highfis import HTSKClassifierEstimator

clf = HTSKClassifierEstimator(
    n_mfs=3,
    rule_base="en",
    epochs=200,
    learning_rate=1e-3,
    ur_weight=0.01,
    random_state=42,
)
clf.fit(X_train, y_train)
print(clf.score(X_test, y_test))

Works with sklearn.pipeline.Pipeline, GridSearchCV, and cross_val_score.

🧩 Key Components

Class	Module	Description
BaseTSK	highfis.base	Abstract base for TSK models with unified training loop.
GaussianMF	highfis.memberships	Differentiable Gaussian membership function.
TriangularMF	highfis.memberships	Triangular membership function.
TrapezoidalMF	highfis.memberships	Trapezoidal membership function.
BellMF	highfis.memberships	Generalized bell membership function.
SigmoidalMF	highfis.memberships	Sigmoidal membership function.
SoftmaxLogDefuzzifier	highfis.defuzzifiers	Stable softmax(log(w)) normalization.
SumBasedDefuzzifier	highfis.defuzzifiers	Classic w / sum(w) normalization.
LogSumDefuzzifier	highfis.defuzzifiers	Temperature-scaled log-space normalization.
MembershipLayer	highfis.layers	Evaluates all membership functions.
RuleLayer	highfis.layers	Computes firing strengths with configurable t-norm and rule base.
ClassificationConsequentLayer	highfis.layers	Linear TSK consequent aggregation for classification.
RegressionConsequentLayer	highfis.layers	Linear TSK consequent aggregation for regression.
HTSKClassifier	highfis.models	Full TSK classification pipeline as nn.Module.
HTSKRegressor	highfis.models	Full TSK regression pipeline as nn.Module.
HTSKClassifierEstimator	highfis.estimators	sklearn-compatible classification estimator.
HTSKRegressorEstimator	highfis.estimators	sklearn-compatible regression estimator.
InputConfig	highfis.estimators	Per-feature membership function configuration.

Structural Typing Protocols

Protocol	Description
MembershipFn	Any callable (Tensor) → Tensor for membership degrees.
TNorm	Any callable (Tensor) → Tensor for rule aggregation.
Defuzzifier	Any callable (Tensor) → Tensor for firing-strength normalization.
ConsequentFn	Any callable (Tensor, Tensor) → Tensor for consequent output.

🧪 Testing & Quality

Running tests

Run the full test suite with coverage:

hatch test -c -a

This project is tested on Python 3.11 | 3.12 | 3.13 | 3.14 across Linux, Windows and macOS.

Linting & Formatting

hatch fmt

Typing

hatch run typing

Security

hatch run security

📚 Documentation

Comprehensive guides, API reference, and examples: dcruzf.github.io/highFIS.

🤝 Contributing

Issues and pull requests are welcome! Please open a discussion if you'd like to propose larger changes.

📄 License

Distributed under the GPLv3.