While multiple types of biases can occur in clinical machine learning, the status quo in algorithmic debiasing is to optimize a single fairness metric in the training procedure. We propose a multi-adversarial debiasing framework that builds on the established ... ...

请注册登录后继续浏览