reconstruct.composed_projector: Reconstruct Data from Scores using a Composed Projector

In bbuchsbaum/multivarious: Extensible Data Structures for Multivariate Analysis

Reconstruct Data from Scores using a Composed Projector

Description

Maps scores from the final latent space back towards the original input space using the composed projector's combined inverse projection. Requires scores to be provided explicitly.

Usage

## S3 method for class 'composed_projector'
reconstruct(x, scores, comp = NULL, rowind = NULL, colind = NULL, ...)

Arguments

x	A composed_projector object.
scores	A numeric matrix of scores (observations x components) in the final latent space of the composed projector.
comp	Numeric vector of component indices (columns of scores, rows of inverse_projection) to use for reconstruction. Defaults to all components.
rowind	Numeric vector of row indices (observations in scores) to reconstruct. Defaults to all rows.
colind	Numeric vector of original variable indices (columns of the final reconstructed matrix) to return. Defaults to all original variables.
...	Additional arguments (currently unused).

Details

Attempts to apply the reverse_transform of the first stage's preprocessor to return data in the original units. If the first stage preprocessor is unavailable or invalid, a warning is issued, and data is returned in the (potentially) preprocessed space of the first stage.

Value

A matrix representing the reconstructed data, ideally in the original data space.

bbuchsbaum/multivarious documentation built on July 16, 2025, 11:04 p.m.