X-ANFIS: Modular ANFIS Library with Classic, Gradient, and Bio-inspired Training Strategies

X-ANFIS is a Python library offering a powerful and extensible implementation of Adaptive Neuro-Fuzzy Inference System (ANFIS) using PyTorch and Scikit-Learn. The library is written with object-oriented principles and modular architecture, enabling easy customization, integration, and experimentation.It supports:

• 🔧 Classic ANFIS (Hybrid: Gradient Descent + Least Squares)
• 🌊 Gradient-based ANFIS (Fully differentiable training)
• 🧠 Bio-inspired ANFIS (Metaheuristics + Least Squares)

✨ Key Features

• Modular, object-oriented design for ease of extension.
• Scikit-Learn API: .fit(), .predict(), .score() and compatible with GridSearchCV, Pipeline, etc.
• Wide range of Gradient-based training optimizers: SGD, Adam, RMSprop, Adagrad, AdamW, ...
• Wide range of Bio-inspired optimizers: GA, PSO, ACO, WOA, BA, FFA, CSA, GWO, ABC, DE, ...
• Rich membership functions: Triangular, Gaussian, Bell, ...
• Over 67 built-in metrics and losses.

🧠 Model Zoo

| Model Class | Training Method | Type | |-------------------------------------------|---------------------|-----------------------------| | AnfisRegressor, AnfisClassifier | Classic Hybrid | Regression / Classification | | GdAnfisRegressor, GdAnfisClassifier | Gradient-based | Regression / Classification | | BioAnfisRegressor, BioAnfisClassifier | Metaheuristic-based | Regression / Classification |

📌 Citation

Please include these citations if you plan to use this library:

```bibtex @software{thieu20250414, author = {Nguyen Van Thieu}, title = {X-ANFIS: An Extensible and Cross-Learning ANFIS Framework for Machine Learning Tasks}, month = June, year = 2025, doi = {10.6084/m9.figshare.28802531}, url = {https://github.com/thieu1995/X-ANFIS} }

@article{van2023mealpy, title = {MEALPY: An open-source library for latest meta-heuristic algorithms in Python}, author = {Van Thieu, Nguyen and Mirjalili, Seyedali}, journal = {Journal of Systems Architecture}, year = {2023}, publisher = {Elsevier}, doi = {10.1016/j.sysarc.2023.102871} } ```

⚙️ General tutorial

Below tutorial is how you can install and use this library. For more complex examples and documentation please check the examples folder and documentation website.

📦 Installation

Install the latest version using pip:

bash pip install xanfis

After that, check the version to ensure successful installation:

```bash $ python

import xanfis xanfis.version ```

🧪 Quick Example

Let's say I want to use Adam optimization-based ANFIS for Iris classification dataset. Here how to do it.

```python from xanfis import Data, GdAnfisClassifier from sklearn.datasets import load_iris

Load data object

X, y = loadiris(returnX_y=True) data = Data(X, y)

Split train and test

data.splittraintest(testsize=0.2, randomstate=2, inplace=True, shuffle=True) print(data.Xtrain.shape, data.Xtest.shape)

Scaling dataset

data.Xtrain, scalerX = data.scale(data.Xtrain, scalingmethods=("standard", "minmax")) data.Xtest = scalerX.transform(data.X_test)

data.ytrain, scalery = data.encodelabel(data.ytrain) data.ytest = scalery.transform(data.y_test)

print(type(data.Xtrain), type(data.ytrain))

Create model

model = GdAnfisClassifier(numrules=20, mfclass="Gaussian", actoutput=None, vanishingstrategy="blend", reglambda=None, epochs=50, batchsize=16, optim="Adam", optimparams={"lr": 0.01}, earlystopping=True, npatience=10, epsilon=0.001, validrate=0.1, seed=42, verbose=True)

Train the model

model.fit(X=data.Xtrain, y=data.ytrain)

Test the model

ypred = model.predict(data.Xtest) print(ypred) print(model.predictproba(data.X_test))

Calculate some metrics

print(model.evaluate(ytrue=data.ytest, ypred=ypred, list_metrics=["F2S", "CKS", "FBS", "PS", "RS", "NPV", "F1S"])) ```

💬 Contributing

We welcome contributions to X-ANFIS! If you have suggestions, improvements, or bug fixes, feel free to fork the repository, create a pull request, or open an issue.

X-ANFIS allows for customization at various levels, including:

• Membership Functions: You can define custom membership functions to fit your problem requirements.

• Learning Strategies: Easily switch between gradient-based or bio-inspired algorithms for training.

• Model Components: Customize the architecture of the ANFIS model, including the number of input and output nodes, output activation function, number of rules, and rule strengths, L2 regularization, training methods.

📞 Community & Support

• 📖 Official Source Code
• 📖 Official Releases
• 📖 Official Docs
• 💬 Telegram Chat
• 🐛 Report Issues
• 🔄 Changelog

🧩 Related Projects

Explore other projects by the author:

• 🔧 MEALPY
• 🔍 Metaheuristics
• 🧪 Permetrics
• 📦 Opfunu
• 🔬 PFEvaluator
• 🧠 IntelELM
• 🔥 Reflame
• 🧭 MetaCluster
• 🧠 Enoppy
• 🤖 AIIR Team

Developed by: Thieu @ 2025