highfis 0.8.0

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

highFIS is a PyTorch-based framework for high-dimensional Takagi–Sugeno–Kang (TSK) fuzzy systems. It brings differentiable fuzzy inference, numerical stability, and sklearn-compatible estimators to both classification and regression workflows. It also includes DGTSK dynamic-gating models for feature and rule selection in high-dimensional fuzzy systems.

🚀 Quick Start

Install from PyPI:

pip install highfis

Run a classifier:

from highfis import HTSKClassifierEstimator

clf = HTSKClassifierEstimator(
    n_mfs=4,
    mf_init="kmeans",
    epochs=150,
    learning_rate=1e-3,
    random_state=42,
)
clf.fit(X_train, y_train)
print(f"Test accuracy: {clf.score(X_test, y_test):.4f}")

highFIS works with sklearn.pipeline.Pipeline, GridSearchCV, and cross_val_score.

Models Available

highFIS offers a family of TSK fuzzy models optimized for different high-dimensional behaviors.

• Vanilla TSK — the original Takagi-Sugeno-Kang model with Gaussian MFs, product antecedent aggregation, and sum-based normalization.
• HTSK — high-dimensional TSK with geometric mean aggregation and log-space normalization to reduce dimensionality bias.
• LogTSK — log-space defuzzification with temperature scaling for numerically stable high-dimensional aggregation.
• AYATSK — adaptive Yager aggregation with sum-based consequent normalization.
• DombiTSK — Dombi t-norm aggregation with first-order consequents and a learnable shape parameter.
• AdaTSK — adaptive Dombi inference using Composite Gaussian MFs with a positive lower bound.
• FSRE-AdaTSK — adaptive model with gated feature selection and rule extraction in the consequent.
• DG-TSK — double-gated training for simultaneous feature selection and rule extraction, followed by first-order fine tuning.
• DG-ALETSK — DG-based adaptive Ln-Exp softmin with embedded feature and rule gates for sparse high-dimensional modeling.

🔧 What’s Included

Core models

• HTSKClassifier, HTSKRegressor
• TSKClassifier, TSKRegressor
• DombiTSKClassifier, DombiTSKRegressor
• AYATSKClassifier, AYATSKRegressor
• AdaTSKClassifier, AdaTSKRegressor
• DGALETSKClassifier, DGALETSKRegressor
• DGTSKClassifier, DGTSKRegressor
• FSREAdaTSKClassifier, FSREAdaTSKRegressor
• LogTSKClassifier, LogTSKRegressor

Estimator wrappers

• HTSKClassifierEstimator, HTSKRegressorEstimator
• TSKClassifierEstimator, TSKRegressorEstimator
• LogTSKClassifierEstimator, LogTSKRegressorEstimator
• AdaTSKClassifierEstimator, AdaTSKRegressorEstimator
• AYATSKClassifierEstimator, AYATSKRegressorEstimator
• DGALETSKClassifierEstimator, DGALETSKRegressorEstimator
• DGTSKClassifierEstimator, DGTSKRegressorEstimator
• FSREAdaTSKClassifierEstimator, FSREAdaTSKRegressorEstimator

Building blocks

• Memberships: GaussianMF, TriangularMF, TrapezoidalMF, BellMF, SigmoidalMF, DiffSigmoidalMF, ProdSigmoidalMF, SShapedMF, LinSShapedMF, ZShapedMF, LinZShapedMF, PiMF, CompositeExponentialMF, GaussianPIMF
• Defuzzifiers: SoftmaxLogDefuzzifier, SumBasedDefuzzifier, LogSumDefuzzifier
• T-norms: prod, min, gmean, dombi, yager, yager_simple, ale_softmin_yager
• Rule base strategies: cartesian, coco, en, custom
• Persistence: estimator .save(path) / .load(path) and versioned checkpoint validation via highfis.persistence

📚 Documentation

The published documentation is available at https://dcruzf.github.io/highFIS.

Key reference pages:

• TSK Vanilla
• LogTSK
• AYATSK
• HTSK
• DombiTSK
• AdaTSK
• DGTSK
• DG-ALETSK
• FSRE-AdaTSK
• Persistence API

🤝 Contributing & Development

See the published contribution guide at contributing.

📄 License

Distributed under the GPLv3.