reparameterize

behavenet.models.vaes.reparameterize(mu, logvar)

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_frames, n_latents)

Return type:

torch.Tensor