Column-wise k-fold cross-validation for nmfkc.signed

Column-wise k-fold CV by held-out samples: for each fold, the model is fit on the training columns and evaluated on the held-out columns by solving for new-sample coefficients via a weighted refit with \(X\) fixed.

Usage

nmfkc.signed.cv(Y, A, rank = 2, ...)

Arguments

Y: Real-valued \(Q_{\mathrm{obs}} \times N\) response matrix (signed entries allowed).
A: Real-valued \(D \times N\) covariate matrix (signed).
rank: Integer \(Q\).
...: Passed to nmfkc.signed; also accepts nfolds (default 5; div alias), seed (default 123), shuffle (default TRUE).

Value

A list with objfunc (mean squared prediction error), sigma (RMSE), objfunc.block (per-fold MSE vector), block (integer fold assignment of length \(N\)). Field names match nmfkc.cv.

Lifecycle

This function is experimental.

References

Ding, C. H. Q., Li, T., & Jordan, M. I. (2010). Convex and semi-nonnegative matrix factorizations. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(1), 45–55.