Skip to main content
. Author manuscript; available in PMC: 2016 Aug 10.
Published in final edited form as: J Am Stat Assoc. 2012 Jan 24;106(495):1009–1024. doi: 10.1198/jasa.2011.tm10332

Algorithm.

The cICA–YW Algorithm

Initialize , λ~.
While the Amari error (9) is greater than the convergence threshold,
    1. Estimate the sources by = W̃X. For j = 1, . . . ,M,
        (a) Estimate ϕ~j using the Yule–Walker method with the order j selected by AIC.
        (b) Compute σ~j2 according to (7).
    2. Update and λ~ via the Newton–Raphson updating (8), using the estimates ϕ~ and σ~2.