Reproduce “Rethinking Learning from Label Proportions via Moment Matching” 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
Tianhao Ma, Wei Wang, Yivan Zhang, Dong-Dong Wu, Takashi Ishida, Gang Niu, Masashi Sugiyama
LLP-MM uses higher-order moment matching to predict individual labels from bag-level class proportions. It treats proportions as information about combinations of distinct labels, assuming labels are conditionally independent given instances. Across synthetic tests on three datasets and three bag strategies, it ranks first in average accuracy in all nine settings, but this does not establish classifier consistency or real-world performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.