Reproduce “Stimulus symmetries can confound representational similarity analyses” 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
Farhad Pashakhanloo, Jacob A. Zavatone-Veth
The authors show how stimulus symmetries can make neural codes with symmetry-respecting function appear different under representational similarity matrices (RSMs). They analyze receptive fields and test a two-layer autoencoder trained on rotations of one character image. Across networks, normalized RSM similarity tracks inferred gauge phase, while reconstruction loss shows no clear relation; this result does not prove identical function.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.