nmfkc.residual.plot: Plot Diagnostics: Original, Fitted, and Residual Matrices as...

In nmfkc: Non-Negative Matrix Factorization with Kernel Covariates

Plot Diagnostics: Original, Fitted, and Residual Matrices as Heatmaps

Description

This function generates a side-by-side plot of three heatmaps: the original observation matrix Y, the fitted matrix XB (from NMF), and the residual matrix E (Y - XB). This visualization aids in diagnosing whether the chosen rank Q is adequate by assessing if the residual matrix E appears to be random noise.

The axis labels (X-axis: Samples, Y-axis: Features) are integrated into the main title of each plot to maximize the plot area, reflecting the compact layout settings.

Usage

nmfkc.residual.plot(
  Y,
  result,
  fitted.palette = (grDevices::colorRampPalette(c("white", "orange", "red")))(256),
  residual.palette = (grDevices::colorRampPalette(c("blue", "white", "red")))(256),
  ...
)

Arguments

Y	The original observation matrix (P x N).
result	The result object returned by the nmfkc function.
fitted.palette	A vector of colors for Y and XB heatmaps. Defaults to white-orange-red. For backward compatibility, Y_XB_palette is accepted via ....
residual.palette	A vector of colors for the residuals heatmap. Defaults to blue-white-red. For backward compatibility, E_palette is accepted via ....
...	Additional graphical parameters passed to the internal image calls.

Value

NULL. The function generates a plot.

See Also

nmfkc, residuals.nmf

Examples

Y <- t(iris[1:30, 1:4])
result <- nmfkc(Y, rank = 2)
nmfkc.residual.plot(Y, result)

nmfkc documentation built on July 14, 2026, 1:07 a.m.