ANFIS Toolbox

ANFIS Toolbox is a Python library for creating, training and deploying Adaptive Neuro-Fuzzy Inference Systems. It implements the classic Takagi-Sugeno-Kang ANFIS on top of NumPy, behind an interface that will look familiar to anyone who has used scikit-learn.

pip install anfis-toolbox

import micropip
await micropip.install("anfis-toolbox")  # only needed here, in the browser

import numpy as np
from anfis_toolbox import ANFISRegressor

rng = np.random.default_rng(0)
X = rng.uniform(-2, 2, (200, 2))
y = X[:, 0] ** 2 + X[:, 1] ** 2

model = ANFISRegressor()
model.fit(X, y)
print("R2:", round(model.score(X, y), 4))

What is in it

• Two estimators, ANFISRegressor and ANFISClassifier, with fit, predict, score and an evaluate that reports the usual metrics.
• The four-layer architecture: membership, rules, normalization and consequent. How ANFIS works walks through it.
• Thirteen membership functions, from Gaussian and bell shapes to triangular, trapezoidal, sigmoidal, S, Z and Pi shapes.
• Trainers kept apart from the model: hybrid (gradient descent plus least squares), SGD, Adam, RMSProp and particle swarm optimization.
• Initialization strategies for the fuzzy sets: fuzzy c-means, grid and random.

Articles

• How ANFIS works
• A first model with ANFIS Toolbox

Links