Reproduce “Learning from a Complementary-label Source Domain: Theory and Algorithms” 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
Yiyang Zhang, Feng Liu, Zhen Fang, Bo Yuan, Guangquan Zhang, Jie Lu
CLARINET jointly trains source classification from complementary labels and adaptation to unlabeled target data. It sharpens class predictions for conditional domain alignment, without first assigning source pseudo-labels. Across seven tested image-transfer tasks, mean target-test accuracy is 77.377%, versus 74.400% for the strongest two-step baseline. Labels were uniformly selected incorrect classes, limiting evidence to that setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.