Multivariate Gaussianization for Data Processing
Multivariate Gaussianization for Data Processing
Multivariate Gaussianization for Data Processing
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Intro Iterative <strong>Gaussianization</strong> Experiments ConclusionsExperiment 1: Density estimation toy examplesDensity estimation with RBIG1: Input: Given data x (0) = [x 1, . . . , x d ] ⊤ ∈ R d2: Learn the sequence of <strong>Gaussianization</strong> trans<strong>for</strong>ms, G, such that y = G(x)3: Compute its Jacobian, J G4: The p y(y) is a multivariate Gaussian:(1p y(y) = p y(G(x)) =( √ 2π|Σ|) exp − 1 )d 2 (G(x) − µ y) ⊤ Σ −1 (G(x) − µ y )5: Compute the probability in the input space with:p x(x) = p y(y) · |∇ xG|