plot.nmfre: Plot convergence diagnostics for NMF models

In nmfkc: Non-Negative Matrix Factorization with Kernel Covariates

Plot convergence diagnostics for NMF models

Description

Plots the objective function value over iterations for nmfre and nmf.ffb objects. (For nmfkc and nmfae, plot methods are defined in their respective source files.)

Usage

## S3 method for class 'nmfre'
plot(x, ...)

## S3 method for class 'nmf.ffb'
plot(x, ..., which = c("penalized", "reconstruction", "both"))

Arguments

x	A fitted model object.
...	Additional graphical arguments passed to plot.
which	For plot.nmf.ffb: which objective to plot. One of "full" (default; loss + penalties, the actual monotonically-decreasing quantity that the multiplicative updates minimize), "reconstruction" (Frobenius distance only, \| Y_1 - X B \|_F^2), or "both" (overlay both with a legend). "both" is useful for diagnosing whether regularization is actively shaping the solution: if the two curves diverge, the penalties are pulling the optimizer away from the pure least-squares minimum. Ignored for nmf.ffb(method = "fiml") objects, which carry no iteration trace: for them the BIC of the L1 path (x$path) is drawn against log(lambda1): the line follows the smallest BIC at each penalty (over the starts), every individual proposal is a small grey point, the number of selected paths is printed at each point and the BIC-selected penalty is circled.

Value

Invisible NULL.

See Also

nmfre, nmf.ffb

Examples

set.seed(1)
Y <- matrix(runif(20), nrow = 4)
A <- diag(5)
res <- nmfre(Y, A, rank = 2)
plot(res)

nmfkc documentation built on Oct. 30, 2026, 9:06 a.m.