Skeptical inferences in multi-label ranking with sets of probabilities

TL;DR

This paper addresses multi-label ranking under uncertainty by using credal sets to produce set-valued predictions, offering a skeptical inference approach that accounts for imprecise probabilities.

Contribution

It introduces a novel framework for skeptical inference in multi-label ranking using convex sets of probabilities, moving beyond single predictions.

Findings

Develops a credal set-based method for multi-label ranking

Enables set-valued predictions reflecting uncertainty

Provides a new perspective on skeptical inference in ranking

Abstract

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.