Matrix completion using nuclear norm, spectral norm or weighted nuclear norm minimization.

Completes a matrix with missing entries, such that the obtained matrix has minimal norm.

Completes a matrix with missing entries, such that the obtained matrix has minimal norm.
Usage:
[CompletedMat, ier] = MatrixCompletion(A.*B, B,N, mode, lambda_tol, tol, display);
A - matrix to complete
B - binary matrix indicates values or missing entries in A (same size, 1 - known value, 0 - missing value)
N - number of iterations
mode - working mode: can be 'nuclear', 'spectral'
lambda_tol - tolerance value from the nuclear/spectral norm minimal value
tol - tolerance on the known entries
Output:
CompletedMat - completed matrix
ier - error indicator: 0 - OK, 1 - Failed to converge (probably needs more iterations).
For a demonstration, run demo.m
The code is based on the paper:
G. Shabat, A. Averbuch "Interest Zone Matrix Approximation", Electronic journal of Linear Algebra, Vol. 23(1), 2012.

Note: The package contains other useful code such as approximating a matrix under different constraints by taking all of it into account or just part of its entries.