Matrix completion based on Gaussian parameterized belief propagation

Koki Okajima, Yoshiyuki Kabashima

We develop a message-passing algorithm for noisy matrix completion problems based on matrix factorization. The algorithm is derived by approximating message distributions of belief…