Variational Autoencoders
Why
Definition
A variational autoencoder
(VAE) from
latent set to
observation set is an
ordered pair whose first
coordinate is a deep latent generation pair from
to (with parameters ) and whose
second coordinate is a deep conditional
distribution from to (with parameters
).
A VAE inherits its joint
function from its deep latent generation
pair.
is called the
latent distribution (or
prior distribution,
latent model).
is called the
decoder distribution.
is called the
encoder distribution (or
inference distribution,
recognition distribution).
A variational autoencoder
family, from to , is a family of
autoencoders .