Regression¶
Predict, classify, and explain with functional predictors and responses.
The Regression module covers the full spectrum of supervised learning with functional data -- from classical scalar-on-function models to elastic regression, classification, conformal prediction, and model explainability.
-
:material-chart-line:{ .lg .middle } Scalar-on-Function
Predict a scalar response from functional predictors using FPC regression, PLS, nonparametric kernel methods, and automatic model selection.
-
:material-chart-bell-curve:{ .lg .middle } Function-on-Scalar
Model a functional response as a function of scalar predictors via FOSR and functional ANOVA.
-
:material-label:{ .lg .middle } Classification
Classify functional observations using LDA, QDA, k-NN, kernel methods, functional logistic regression, and cross-validated model comparison.
-
:material-swap-horizontal:{ .lg .middle } Elastic Regression
Phase-invariant scalar-on-function regression and logistic regression under the elastic metric. Jointly aligns curves and estimates the regression model.
-
:material-lightbulb-on:{ .lg .middle } Explainability
Interpret functional regression models via permutation importance, partial dependence plots, SHAP values, beta decomposition, and significant region detection.
-
:material-shield-check:{ .lg .middle } Conformal Prediction
Distribution-free prediction intervals and prediction sets with finite-sample coverage guarantees for functional regression and classification.
-
:material-shield:{ .lg .middle } Robust Regression
L1 and Huber M-estimation for functional linear models that resist outlier contamination.