Reproduce “The role of locality and weight sharing in convolutional neural 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

Francesco Mori, Melanie Weber

The authors compare convolutional students that share a kernel or learn position-specific kernels in a Gaussian teacher–student setting. Inputs have independent or uniformly correlated positions, and training uses online gradient descent with noisy labels. For independent positions, CNN and LCN error trajectories match after rescaling LCN training time by P; with correlated inputs, rate tuning can favor the CNN late.

Summary by alphaXiv.

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