Most work in algorithmic fairness to date has focused on discrete outcom...
It is usual to consider data protection and learnability as conflicting
...
We propose a novel hybrid loss for multiclass and structured prediction
...
We present a method for learning max-weight matching predictors in bipar...
share