This a Bayesian ICA algorithm for the linear instantaneous mixing model with additive Gaussian noise [1]. The inference problem is solved by ML-II, i.e. the sources are found by integration over the source posterior and the noise covariance and mixing matrix are found by maximization of the marginal likelihood [1]. The sufficient statistics are estimated by either variational mean field theory with the linear response correction or by adaptive TAP mean field theory [2,3]. The mean field equations are solved by a belief propagation method [4] or sequential iteration. The computational complexity is N M^3, where N is the number of time samples and M the number of sources.
calculatePXTheta---Calculate the probability of each pixel being its color conditioned on all of the clusters that were found at the previous (coarser) iteration.
Rao-Blackwellised Particle Filters (RBPFs) are a class of Particle
Filters (PFs) that exploit conditional dependencies between
parts of the state to estimate. By doing so, RBPFs can
improve the estimation quality while also reducing the overall
computational load in comparison to original PFs. However,
the computational complexity is still too high for many
real-time applications. In this paper, we propose a modified
RBPF that requires a single Kalman Filter (KF) iteration per
input sample. Comparative experiments show that while good
convergence can still be obtained, computational efficiency is
always drastically increased, making this algorithm an option
to consider for real-time implementations.