SONFIS

This module contains the implementation of the SONFIS (Self-Organizing Neuro-Fuzzy Inference System) algorithm, which combines a parameter learning algorithm with structural adaptation operators for rule growing, splitting, and pruning. This algorithm is only applicable to neuro_fuzzy_toolbox.models.anfis.rule_reduced_ANFIS models.

For more details on the SONFIS algorithm, refer to the original paper: SONFIS: Structure Identification and Modeling with a Self-Organizing Neuro-Fuzzy Inference System.

SONFIS algorithm pseudocode.
class neuro_fuzzy_toolbox.training.sonfis.SONFIS(Ngrow, dGrow, Nsplit, eSplit, Nvanish, lVanish, max_iterations, ANFIStrainer, early_stopping=None, lse_for_new_consequents=False, lse_for_new_consequents_lambda=0.0, last_training_iteration=False)[source]

Bases: base_model_trainer

Self-Organizing Neuro-Fuzzy Inference System (SONFIS) algorithm.

Combines a parameter learning algorithm with three structural adaptation operators (GrowNet, SplitSubNet, and VanishNet) to iteratively update both the parameters and the structure of a rule-reduced ANFIS model.

Note

This algorithm is only applicable to instances of rule_reduced_ANFIS.

__call__(ANFISmodel, train_loader, val_loader=None, verbose=True)[source]

Runs the SONFIS algorithm.

At each iteration, the structural adaptation operators GrowNet, SplitSubNet, and VanishNet are applied, followed by a parameter update of the modified subnets. The algorithm stops when no structural updates occur or the maximum number of iterations is reached. If early stopping is enabled and a validation DataLoader is provided, the algorithm may also stop early based on the validation loss.

Parameters:

  • ANFISmodel (rule_reduced_ANFIS) – Rule-reduced ANFIS model to train.

  • train_loader (DataLoader) – DataLoader containing the training data.

  • val_loader (DataLoader) – DataLoader containing the validation data. Defaults to None.

  • verbose (bool) – If True, prints progress and structural update messages at each iteration. Defaults to True.

__init__(Ngrow, dGrow, Nsplit, eSplit, Nvanish, lVanish, max_iterations, ANFIStrainer, early_stopping=None, lse_for_new_consequents=False, lse_for_new_consequents_lambda=0.0, last_training_iteration=False)[source]

Initializes a new SONFIS instance.

Parameters:

  • Ngrow (int) – Minimum number of poorly modeled samples required to grow a new subnet.

  • dGrow (float) – Threshold for identifying poorly modeled samples. A sample is considered poorly modeled if its maximum firing level across all subnets is less than or equal to this value raised to the power of the input dimensionality.

  • Nsplit (int) – Minimum number of samples associated with a subnet for it to be considered for splitting.

  • eSplit (float) – Minimum loss value of the samples associated with a subnet for it to be considered for splitting.

  • Nvanish (int) – Maximum number of samples associated with a subnet below which its age counter is incremented.

  • lVanish (int) – Maximum age of a subnet before it is removed.

  • max_iterations (int) – Maximum number of structural adaptation iterations.

  • ANFIStrainer (base_model_trainer) – Instantiated ANFIS training algorithm that defines how the model parameters are updated at each iteration.

  • early_stopping (EarlyStopping) – Early stopping mechanism to use during the SONFIS iterations. Defaults to None.

  • lse_for_new_consequents (bool) – If True, the consequent parameters of rules generated by GrowNet or SplitSubNet are initialized using least-squares estimation instead of random initialization. Defaults to False.

  • lse_for_new_consequents_lambda (float) – Lambda value for Ridge regularization in the least-squares initialization of new consequent parameters. If 0., no regularization is applied. Defaults to 0..

  • last_training_iteration (bool) – If True, performs a final parameter update over all subnets after the SONFIS algorithm finishes. Defaults to False.