Evaluate an NMF model

Calculate loss for an NMF model using the specified loss function, accounting for any masking schemes requested during fitting.

Usage

evaluate(x, ...)

# S4 method for class 'nmf'
evaluate(
  x,
  data,
  mask = NULL,
  missing_only = FALSE,
  loss = c("mse", "gp"),
  test_fraction = 0,
  test_seed = NULL,
  eval_set = c("all", "test", "train"),
  threads = 0,
  verbose = FALSE,
  ...
)

Arguments

x

fitted model, class nmf, generally the result of calling nmf, with models of equal dimensions as data

...

advanced parameters. See Advanced Parameters section.

data

dense or sparse matrix of features in rows and samples in columns. Prefer matrix or Matrix::dgCMatrix, respectively. Also accepts a file path (character string) which will be auto-loaded based on extension.

mask

missing data mask. Accepts: NULL (no masking), "zeros" (mask zeros), "NA" (mask NAs), a dgCMatrix/matrix (custom mask), or list("zeros", <matrix>) to mask zeros and use a custom mask simultaneously.

missing_only

calculate loss only for missing values specified as a matrix in mask

loss

loss function to use: "mse" (Mean Squared Error, default) or "gp" (Generalized Poisson / KL divergence)

test_fraction

fraction of entries to hold out as test set (default 0 = disabled). When > 0, creates a random mask for test/train split.

test_seed

seed for test set generation. If NULL, attempts to use test mask from model's @misc$test_mask if available.

eval_set

which set to evaluate: "all" (default), "test" (held-out entries only), or "train" (non-held-out entries only). Only used when test_fraction > 0 or test mask exists in model.

threads

number of threads for OpenMP parallelization (default 0 = all available)

verbose

print progress information (default FALSE)

Value

A single numeric value: the loss (MSE or GP/KL divergence) of the model on the data.

See also

nmf

Examples

# \donttest{
library(Matrix)
A <- rsparsematrix(100, 50, 0.1)
model <- nmf(A, 3, seed = 1, maxit = 50, tol = 1e-4)
evaluate(model, A)  # MSE
#> [1] 0.09021044
# }