Reproduce “Learning the Irreducible Representations of Commutative Lie Groups” 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
Taco Cohen, Max Welling
The authors propose a method that learns image features invariant to transformations by modeling pairs of related images and combining their changing responses. Its components are irreducible representations, and its probabilistic model computes a posterior over transformations. On rotated MNIST, nearest-neighbor classification with invariant features outperforms pixel and tangent distances; this evidence applies to compact commutative groups of orthogonal transformations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.