As one of the classical strategies for positive-unlabeled (PU) learning, the cost-sensitive methods achieve binary classification by minimizing the overall risk. To achieve this goal, the class prior is usually exploited to guide model learning in many state-o... ...

请注册登录后继续浏览