Variational Composite Autoencoders
This work addresses inefficiencies in variational autoencoders for researchers in machine learning, though it appears incremental as it builds on existing hierarchical models.
The paper tackles the challenge of learning in latent variable models with complex data structures or intractable latent variables by proposing a variational composite autoencoder that amortizes over a hierarchical latent variable model, and experimental results confirm its advantages.
Learning in the latent variable model is challenging in the presence of the complex data structure or the intractable latent variable. Previous variational autoencoders can be low effective due to the straightforward encoder-decoder structure. In this paper, we propose a variational composite autoencoder to sidestep this issue by amortizing on top of the hierarchical latent variable model. The experimental results confirm the advantages of our model.