Reproduce “Walking the Weight Manifold: a Topological Approach to Conditioning Inspired by Neuromodulation” 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
Ari S. Benjamin, Kyle Daruwalla, Christian Pehle, Abdul-Malik Zekri, Anthony M. Zador
The authors propose a neural network that lets a low-dimensional task condition select weights from a smoothly changing family, such as a line or ellipse. On rotated CIFAR-10 classification, the ellipse model has higher test accuracy than unconditioned and concatenation baselines across tested training-condition sparsities, supporting generalization to unseen rotation angles in this setting; the comparison does not isolate the source of the gain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.