Reproduce “Deep Pyramidal Residual 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
Dongyoon Han, Jiwhan Kim, Junmo Kim
PyramidNet distributes channel growth across residual units instead of concentrating it at downsampling transitions, using zero-padded identity shortcuts when channels increase. It tests a block variant removing the first ReLU and adding batch normalization after the final convolution. The authors report 20.5% ImageNet-1k top-1 error versus 21.7% for pre-activation ResNet-200, but the comparison does not isolate the cause of the difference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.