SVD with Missing Value Handling

Performs SVD on a matrix, iteratively imputing missing values (NaN/NA). Based on the algorithm from misssvd.m (Claus A. Andersson).

Usage

misssvd(X, truncate = TRUE, conv_lim = 1e-12, max_iter = 100)

Arguments

X

Matrix to decompose

truncate

Logical; if TRUE, truncate to rank of X (economy SVD)

conv_lim

Convergence limit for missing value iteration

max_iter

Maximum iterations for missing value imputation

Value

List with components:

u

Left singular vectors

d

Singular values (as vector)

v

Right singular vectors

Examples

X <- matrix(rnorm(100), 10, 10)
X[sample(100, 5)] <- NA
result <- misssvd(X)