GrafoRVFL (GRAdient Free Optimized Random Vector Functional Link)

Overview

GrafoRVFL is an open-source Python library designed to optimize Random Vector Functional Link (RVFL) networks using various gradient-free metaheuristic algorithms such as GA, PSO, WOA, TLO, DE, etc. It is fully implemented in NumPy and seamlessly integrates with the Scikit-Learn interface, making it easy to plug into standard ML workflows. GrafoRVFL enables hyperparameter tuning for RVFL networks without relying on gradient-based methods.

Features

• Free software under GNU GPL v3
• Full documentation: https://graforvfl.readthedocs.io
• Estimators:
  • RvflRegressor
  • RvflClassifier
  • GfoRvflCV
  • GfoRvflTuner
  • GfoRvflComparator
• Python compatibility: >= 3.8
• Dependencies:
  • numpy, scipy, scikit-learn, pandas, mealpy, permetrics, matplotlib

Citation Request

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

```bibtex @software{nguyenvanthieu202310258280, author = {Nguyen Van Thieu}, title = {GrafoRVFL: A Gradient-Free Optimization Framework for Boosting Random Vector Functional Link Network}, month = June, year = 2025, publisher = {Zenodo}, doi = {10.5281/zenodo.10258280}, url = {https://github.com/thieu1995/GrafoRVFL} }

@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} }

@inproceedings{nguyen2019building, title={Building resource auto-scaler with functional-link neural network and adaptive bacterial foraging optimization}, author={Nguyen, Thieu and Nguyen, Binh Minh and Nguyen, Giang}, booktitle={International Conference on Theory and Applications of Models of Computation}, pages={501--517}, year={2019}, organization={Springer} }

@inproceedings{nguyen2018resource, title={A resource usage prediction system using functional-link and genetic algorithm neural network for multivariate cloud metrics}, author={Nguyen, Thieu and Tran, Nhuan and Nguyen, Binh Minh and Nguyen, Giang}, booktitle={2018 IEEE 11th conference on service-oriented computing and applications (SOCA)}, pages={49--56}, year={2018}, organization={IEEE}, doi={10.1109/SOCA.2018.00014} } ```

• Learn more about Random Vector Functional Link from this paper

• Learn more about on how to use Gradient Free Optimization to fine-tune the hyper-parameter of RVFL networks from this paper

Installation

Install the latest version from PyPI:

bash $ pip install graforvfl

Verify installation:

```bash $ python

import graforvfl graforvfl.version ```

Example Usage

Below is a simple example code of how to use Gradient Free Optimization to tune hyper-parameter of RVFL network.

```python from sklearn.datasets import loadbreastcancer from graforvfl import Data, GfoRvflCV, StringVar, IntegerVar, FloatVar

Load data object

X, y = loadbreastcancer(returnXy=True) data = Data(X, y)

Split train and test

data.splittraintest(testsize=0.2, randomstate=2, inplace=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)

Design the boundary (parameters)

mybounds = [ IntegerVar(lb=3, ub=50, name="sizehidden"), StringVar(validsets=("none", "relu", "leakyrelu", "celu", "prelu", "gelu", "elu", "selu", "rrelu", "tanh", "hardtanh", "sigmoid", "hardsigmoid", "logsigmoid", "silu", "swish", "hardswish", "softplus", "mish", "softsign", "tanhshrink", "softshrink", "hardshrink", "softmin", "softmax", "logsoftmax"), name="actname"), StringVar(validsets=("orthogonal", "heuniform", "henormal", "glorotuniform", "glorotnormal", "lecununiform", "lecunnormal", "randomuniform", "randomnormal"), name="weightinitializer"), FloatVar(lb=0, ub=10., name="regalpha"), ]

model = GfoRvflCV(problemtype="classification", bounds=mybounds, optim="OriginalWOA", optimparams={"name": "WOA", "epoch": 10, "popsize": 20}, scoring="AS", cv=3, seed=42, verbose=True) model.fit(data.Xtrain, data.ytrain) print(model.bestparams) print(model.bestestimator) print(model.bestestimator.scores(data.Xtest, data.ytest, listmetrics=("PS", "RS", "NPV", "F1S", "F2S"))) ```

The more complicated cases in the folder: examples. You can also read the documentation for more detailed installation instructions, explanations, and examples.

Official channels

• Official source code repository
• Official document
• Download releases
• Issue tracker
• Notable changes log
• Official discussion group

Developed by: Thieu @ 2025