Reproduce “Topographic VAEs learn Equivariant Capsules” 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
T. Anderson Keller, Max Welling
The authors introduce the Topographic VAE, which uses sequences of transformed inputs to encourage changes to correspond to cyclic shifts within groups of latent features called capsules. This variational model represents topographic Student-t variables through Gaussian variables. On dSprites, its L=S/3 setting reports CapCorr of 1.00 ± 0.00 for four factors; equivariance is approximate, not guaranteed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.