ConeRANK: Ranking as Learning Generalized Inequalities
This addresses ranking problems in data mining, but appears incremental as it builds on existing pairwise learning-to-rank methods with a cone-based formulation.
The paper tackles document ranking by formulating it as learning proper cones to establish partial orderings between vectors, proposing the ConeRank algorithm that learns a non-negative subspace as a polyhedral cone. Experimental results on the LETOR 4.0 dataset show ConeRank is competitive against other recent ranking approaches.
We propose a new data mining approach in ranking documents based on the concept of cone-based generalized inequalities between vectors. A partial ordering between two vectors is made with respect to a proper cone and thus learning the preferences is formulated as learning proper cones. A pairwise learning-to-rank algorithm (ConeRank) is proposed to learn a non-negative subspace, formulated as a polyhedral cone, over document-pair differences. The algorithm is regularized by controlling the `volume' of the cone. The experimental studies on the latest and largest ranking dataset LETOR 4.0 shows that ConeRank is competitive against other recent ranking approaches.