Reproduce “Efficient Optimization for Rank-based Loss Functions” 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
Pritish Mohapatra, Michal Rolinek, C.V. Jawahar, Vladimir Kolmogorov, M. Pawan Kumar
The authors introduce a divide-and-conquer method for loss-augmented inference in ranking-based detection training. It partitions negative samples and uses monotonicity in optimal interleaving ranks to avoid sorting the negative set. On PASCAL VOC 2007, mean AP was 36.616 with AP-loss training versus 29.4995 with 0–1 loss; inference took 0.5214 seconds versus 7.623, not total training time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.