Reproduce “Subspace Clifford 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

Zhongping Ji

Subspace Clifford Networks replace CliffordNet’s fixed cyclic shifts with groups that compute within-group wedge interactions and coordinate-wise products, while projections mix channels across groups. This reduces interaction width while retaining dense projections. On the hierarchical CIFAR-100 model, k=3 reaches 80.12% Top-1 accuracy versus 79.53%, with fewer parameters and shorter training time; the comparison does not isolate grouping or establish broader results.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.