Transform New Data Using Existing SVD Compression

Applies a previously computed SVD compression to new data, ensuring consistency across multiple runs or when adding new voxels.

Usage

transform_new_data_svd(new_data, compression_result)

Arguments

new_data

Numeric matrix (N_new x T) with same number of columns as the original training data.

compression_result

List returned by compress_features_svd containing the rotation matrix and scaling parameters.

Value

Compressed and normalized feature matrix (N_new x M) where M is the number of components in the original compression.

Examples

# Train on subset of data
train_data <- matrix(rnorm(500 * 300), 500, 300)
compression <- compress_features_svd(train_data, n_components = 15)
#> Requested n_components (15) explained 13.9% variance; increased to 223 components to retain 95%.

# Apply to new data
test_data <- matrix(rnorm(100 * 300), 100, 300)
compressed_test <- transform_new_data_svd(test_data, compression)