Reproduce “Cost Sensitive Learning of Deep Feature Representations from Imbalanced Data” 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
Salman H. Khan, Munawar Hayat, Mohammed Bennamoun, Ferdous Sohel, Roberto Togneri
The CoSen CNN learns class-dependent costs during training to adjust how class-specific mistakes shape its scores, without changing predictions at inference or resampling examples. On the five-class DIL skin-lesion experiment, its accuracy was 80.2%, compared with 75.2% for the baseline, under three-fold cross-validation. This result applies to that experiment and does not establish clinical benefit.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.