Improving Variational Autoencoders with Inverse Autoregressive Flow
NIPS, pp. 4736-4744, 2016.
We propose a simple and scalable method for improving the flexibility of variational inference through a transformation with autoregressive neural networks. Autoregressive neural networks, such as RNNs or the PixelCNN, are very powerful models and potentially interesting for use as variational posterior approximation. However, ancestral s...More
Full Text (Upload PDF)
PPT (Upload PPT)