Reproduce “Batch-Shaping for Learning Conditional Channel Gated 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
Babak Ehteshami Bejnordi, Tijmen Blankevoort, Max Welling
The authors introduce a residual network that gates individual convolutional channels per image, using a lightweight gate module driven by block inputs. Batch-shaping encourages conditional gates, and a delayed sparsity penalty reduces computation. On ImageNet, ResNet34-BAS achieves 72.55% top-1 accuracy versus 69.76% for ResNet18 at similar average MACs, but this does not establish lower real-world latency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.