Local Models API¶
kfc_procedure.core.ml.base.BaseLocalModel
¶
Bases: BaseEstimator, ABC
Unified base class for all local models in the F-step.
A local model learns from a subset of data (e.g. cluster-wise or divergence-specific partition) and produces predictions that are later combined in the C-step.
This class is intentionally task-agnostic.
fit
abstractmethod
¶
Fit the local model.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
X
|
ndarray
|
Input features. |
required |
y
|
ndarray
|
Target values. |
required |
Returns:
| Type | Description |
|---|---|
self
|
|
predict
abstractmethod
¶
Predict outputs for F-step.
Returns predictions that will be stacked into a prediction matrix for the C-step.
kfc_procedure.core.ml.base.LocalModelFactory
¶
Bases: BaseFactory
Factory for all local models used in F-step.
Supports both regression and classification models via categories:
- regression
- classification
- multitask (optional extension)
kfc_procedure.core.ml.sklearn.MeanRegressor
¶
Bases: BaseLocalModel
kfc_procedure.core.ml.sklearn.SklearnLocalModel
¶
kfc_procedure.core.ml.sklearn.clean_sklearn_name
¶
Convert sklearn estimator name to snake_case.
kfc_procedure.core.ml.sklearn.register_all_sklearn_models
¶
Auto-register sklearn models into LocalModelFactory.