Data dimensionality
Evaluate regularization term for a Gaussian component
Evaluate regularization term for a Gaussian component
Mixture component
regularization value
Evaluate regularization term for the weights vector
Evaluate regularization term for the weights vector
model weights vector
regularization value
Computes the loss function's gradient w.r.t a Gaussian component's parameters
Computes the loss function's gradient w.r.t a Gaussian component's parameters
Mixture component
gradient
Block matrix of prior parameters [A B; C D].
Block matrix of prior parameters [A B; C D]. The blocks are:
A = iwMean + kappa * normalMean * normalMean.t
B = kappa * normalMean
C = kappa * normalMean.t
D = kappa
kappa = degFrredom + dim + 2
set degrees of freedom for the Inverse-Wishart prior
set degrees of freedom for the Inverse-Wishart prior
Degrees of freedom
Set Gaussian parameters' prior means.
Set Gaussian parameters' prior means. The Gaussian parameter prior means must be set at the same time to check that the dimensions match
Expected value vector for the prior Normal distribution
Expected value matrix for the prior Inverse-Wishart distribution
Computes the loss function's gradient with respect to the current weights vector
Computes the loss function's gradient with respect to the current weights vector
current weights vector
gradient
Implementation of conjugate prior regularization; this means an Inverse-Wishart prior over the covariance matrices, a Normal prior over the means and a Dirichlet distribution prior over the weights.