API Reference¶
This section documents the public surface of ANFIS Toolbox. Use it alongside the user guides and examples when you need precise signatures, parameters, and return types.
Estimators¶
ANFISRegressor– Scikit-learn style interface for Takagi–Sugeno–Kang regression.ANFISClassifier– Classification counterpart with probability predictions and evaluation helpers.
Membership Functions¶
Thirteen membership function families covering Gaussian, bell, sigmoidal, and
piecewise-linear shapes are documented in
membership-functions.md. Each entry includes
parameters, derivative support, and usage examples.
Training¶
- Optimizers – Gradient-based, hybrid, and swarm trainers are described in
optim.mdwith configuration notes and supported hyper-parameters. - Losses – Regression and classification objectives (and their gradients)
are listed in
losses.md.
Metrics¶
Evaluation helpers for regression, classification, and clustering are grouped in
metrics.md. Each function documents expected inputs and output
formats so you can integrate metrics into experiments or monitoring.
Core Internals¶
- Models – Low-level ANFIS graph classes and their rule representations live
in
models.md. - Layers – Individual computational layers and their tensor operations are
explained in
layers.md.
These pages are useful when you need to inspect or extend the internal pipeline that powers the high-level estimators.
Utilities¶
- Configuration – Utilities for persisting and replaying setups appear in
config.md. - Logging – Structured training logs and logging configuration are covered
in
logging.md.
Advanced Topics¶
- Builders – Advanced model construction hooks are described in
builders.md. Most users can rely on estimator defaults. - Clustering – The fuzzy C-means implementation used for membership
initialization is detailed in
clustering.md.
Where to Start¶
- New to ANFIS Toolbox? Begin with the models overview.
- Looking for ready-to-run notebooks? Browse the Examples section in the navigation.
- Exploring code while reading docs? The “View source” actions in each page jump straight to the implementation.