vae_loss_correlated: A custom loss function for a VAE learning a multivariate...

In ML2Pvae: Variational Autoencoder Models for IRT Parameter Estimation

A custom loss function for a VAE learning a multivariate normal distribution with a full covariance matrix

Description

A custom loss function for a VAE learning a multivariate normal distribution with a full covariance matrix

Usage

vae_loss_correlated(
  encoder,
  inv_skill_cov,
  det_skill_cov,
  skill_mean,
  kl_weight,
  rec_dim
)

Arguments

encoder	the encoder model of the VAE, used to obtain z_mean and z_log_cholesky from inputs
inv_skill_cov	a constant tensor matrix of the inverse of the covariance matrix being learned
det_skill_cov	a constant tensor scalar representing the determinant of the covariance matrix being learned
skill_mean	a constant tensor vector representing the means of the latent skills being learned
kl_weight	weight for the KL divergence term
rec_dim	the number of nodes in the input/output of the VAE

Value

returns a function whose parameters match keras loss format

ML2Pvae documentation built on May 23, 2022, 9:05 a.m.