Skeptical inferences in multi-label ranking with sets of probabilities
This addresses uncertainty handling in multi-label ranking for applications requiring robust decision-making, but it appears incremental as it adapts existing credal set methods to this specific task.
The paper tackles the problem of making skeptical inferences in multi-label ranking by using convex sets of probabilities (credal sets) to describe uncertainty, resulting in set-valued predictions of completed rankings instead of singleton predictions.
In this paper, we consider the problem of making skeptical inferences for the multi-label ranking problem. We assume that our uncertainty is described by a convex set of probabilities (i.e. a credal set), defined over the set of labels. Instead of learning a singleton prediction (or, a completed ranking over the labels), we thus seek for skeptical inferences in terms of set-valued predictions consisting of completed rankings.