Reproduce “PixelVAE++: Improved PixelVAE with Discrete Prior” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hossein Sadeghi, Evgeny Andriyash, Walter Vinci, Lorenzo Buffoni, Mohammad H. Amin
PixelVAE++ combines an autoregressive PixelCNN++ decoder with three groups of binary latent variables and an RBM prior, reusing decoder downsampling features to help encode one group. On CIFAR-10, the RBM version reaches 2.90 bits per dimension, versus 2.92 with a Gaussian prior, and has higher KL divergence, an indicator of latent information, not proof of semantic structure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.