Multi-prototype Label Ranking with Novel Pairwise-to-Total-Rank Aggregation

We propose a multi-prototype-based algorithm for online learning of soft pairwise-preferences over la-bels. The algorithm learns soft label preferences via minimization of the proposed soft rank-loss measure, and can learn from total orders as well as from various types of partial orders. The soft pair-wise preference algorithm outputs are further ag-gregated to produce a total label ranking prediction using a novel aggregation algorithm that outper-forms existing aggregation solutions. Experiments on synthetic and real-world data demonstrate state-of-the-art performance of the proposed model.