align_pcdirs

The function 'prep_training_data' does not center the warping functions,
which leads to issues when visualizing joint and horizontal principal
component directions. This function aligns the principal directions for
improved interpretability of the principal directions. Currently, only
alignment for jfPCA has been implemented.
The function 'prep_training_data' does not center the warping functions,
which leads to issues when visualizing joint and horizontal principal
component directions. This function aligns the principal directions for
improved interpretability of the principal directions. Currently, only
alignment for jfPCA has been implemented.

Implements the Variable importance Explainable Elastic Shape Analysis pipeline for explainable machine learning with functional data inputs. Converts training and testing data functional inputs to elastic shape analysis principal components that account for vertical and/or horizontal variability. Computes feature importance to identify important principal components and visualizes variability captured by functional principal components. See Goode et al. (2025) <doi:10.48550/arXiv.2501.07602> for technical details about the methodology.

Katherine Goode

veesa

Pipeline for Explainable Machine Learning with Functional Data

J. Derek Tucker

Sandia National Laboratories 

align_pcdirs function

<dl><dt>train_obj</dt>
<dd>Output object from `prep_training_data` (jfpca only)</dd></dl>

Arguments

Obtain PC directions with centered warping functions — align_pcdirs

<dl>

<dt>train_obj</dt>
<dd>Output object from `prep_training_data` (jfpca only)</dd>

</dl>

align_pcdirs: Obtain PC directions with centered warping functions

Description

Usage

Value

Arguments