reparameterize_gaussian¶

daart.models.base.reparameterize_gaussian(mu, logvar)[source]¶

Sample from N(mu, var)

Parameters:

mu (torch.Tensor) – vector of mean parameters
logvar (torch.Tensor) – vector of log variances; only mean field approximation is currently implemented

Returns:

sampled vector of shape (n_sequences, sequence_length, embedding_dim)

Return type:

torch.Tensor