Reconstruct new data in a model's subspace

This function takes a model (e.g., projector or bi_projector) and a new dataset, and computes the rank-d approximation of the new data in the same subspace that was defined by the model. In other words, we project the new data into the fitted subspace and then map it back to the original dimensionality.

Usage

reconstruct_new(x, new_data, ...)

Arguments

x

The fitted model object (e.g., bi_projector) that defines a subspace or factorization.

new_data

A numeric matrix (or data frame) of shape (n x p_full) or possibly fewer columns if you allow partial reconstruction.

...

Additional arguments passed to the specific reconstruct_new method for the class of x.

Value

A numeric matrix (same number of rows as new_data, and typically the same number of columns if you're reconstructing fully) representing the rank-d approximation in the model's subspace.

Details

Similar to reconstruct but operates on an external new_data rather than the original fitted data. Often used to see how well the model's subspace explains unseen data.

See also

reconstruct for reconstructing the original data in the model.

Other reconstruct: reconstruct()