Project new data through a trained factor network

For single-layer results, projects new samples using the trained W and d. For multi-layer results, chains through each layer.

Usage

# S3 method for class 'factor_net_result'
predict(object, newdata, ...)

Arguments

object

A factor_net_result from fit().

newdata

A matrix (features x samples) to project.

...

Additional arguments (currently unused).

Value

For single-layer: a matrix (k x n_new). For multi-layer: a list of H matrices, one per layer.

See also

fit, factor_net