Draw synthetic feature vectors from a trained feature texture model

Samples from the Gaussian copula: draws n_samples from MVN(0, copula_cor), then transforms each feature through its inverse marginal CDF. The result is centered and scaled to zero mean and unit variance per feature, so the texture acts as noise that the structural signal in run_single_iteration() provides the mean for.

Usage

sample_feature_texture(feature_texture, n_samples, n_features = NULL)

Arguments

feature_texture: An iconic_feature_texture object from train_feature_texture().
n_samples: Number of synthetic samples to draw.
n_features: Target number of features. If NULL, uses the number of features in the training data. If larger, additional features are drawn by sampling existing columns with replacement and adding independent noise. If smaller, the first n_features are used.

Value

A n_features x n_samples matrix of synthetic texture values, centered and scaled per feature.

Examples

M <- matrix(rnorm(30 * 200), 30, 200) # 30 transcripts, 200 samples
ft <- train_feature_texture(M)
draws <- sample_feature_texture(ft, 500, n_features = 20)