Reproduce “Deep Learning for Inverse Problems: Bounds and Regularizers” 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
Jaweria Amjad, Zhaoyan Lyu, Miguel R. D. Rodrigues
The authors connect neural network sensitivity to generalization in image reconstruction and propose layer-wise weight orthogonalization as a practical regularizer. Their theoretical bound assumes compact input and observation spaces, an L-Lipschitz forward operator, bounded noise, and IID examples. In ×3 Set14 super-resolution, SRCNN with WO reached 25.33 dB versus 21.94 for weight decay, though SRIP achieved higher PSNR.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.