Reproduce “Equivariant Transformer Networks” 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
Kai Sheng Tai, Peter Bailis, Gregory Valiant
Equivariant Transformers (ETs) help image classifiers handle selected geometric changes by estimating pose in coordinates where a change acts like a shift, then undoing it before classification. On SVHN with 10,000 training examples, an ET-augmented ResNet-10 has 8.24% error, versus 8.73% for plain ResNet-34. This supports data efficiency in that setting, but stacked ETs lack a full-group guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.