That's more related to 'stochastic gradient descent' for 'matrix completion'. The key difference is that Simon Funk's algorithm doesn't treat missing entries as a zero, whereas using linear algebra based techniques on the observed data matrix (formed by putting zeros for unobserved entries) would try to predict the missing entries as zero exactly.
Also related is the 'alternating least squares' algorithm.
Also related is the 'alternating least squares' algorithm.