An Overview of Variational Autoencoders for Source Separation, Finance, and Bio-Signal Applications

. 2021 Dec 28;24(1):55. doi: 10.3390/e24010055

Algorithm 2: AEVB algorithm using minibatches

$θ, ϕ \leftarrow Initialize parameters$
for $do #$ of training iterations:
$X^{(M)} \leftarrow Random minibatch of M datapoints from full dataset X$
$Sample ϵ \sim p (ϵ)$
$Calculate minibatch estimate {\tilde{L}}^{(M)} (θ, ϕ; X^{(M)}, ϵ)$
$Calculate gradients of minibatch estimator \nabla_{θ, ϕ} {\tilde{L}}^{(M)} (θ, ϕ; X^{(M)}, ϵ)$
$θ, ϕ \leftarrow Update parameters using gradients with methods like SGD or Adagrad$
end for