Ladder Variational Autoencoders
This addresses a key bottleneck in unsupervised learning for researchers and practitioners by enabling more effective training of highly expressive deep generative models, though it is an incremental improvement building on prior work like Ladder Networks.
The authors tackled the difficulty of training deep variational autoencoders with multiple stochastic layers by proposing the Ladder Variational Autoencoder, which recursively corrects the generative distribution using a data-dependent approximate likelihood, achieving state-of-the-art predictive log-likelihood and tighter lower bounds compared to existing models.
Variational Autoencoders are powerful models for unsupervised learning. However deep models with several layers of dependent stochastic variables are difficult to train which limits the improvements obtained using these highly expressive models. We propose a new inference model, the Ladder Variational Autoencoder, that recursively corrects the generative distribution by a data dependent approximate likelihood in a process resembling the recently proposed Ladder Network. We show that this model provides state of the art predictive log-likelihood and tighter log-likelihood lower bound compared to the purely bottom-up inference in layered Variational Autoencoders and other generative models. We provide a detailed analysis of the learned hierarchical latent representation and show that our new inference model is qualitatively different and utilizes a deeper more distributed hierarchy of latent variables. Finally, we observe that batch normalization and deterministic warm-up (gradually turning on the KL-term) are crucial for training variational models with many stochastic layers.