Query-Level Stability and Generalization in Learning to Rank

This paper is concerned with the generalization ability of learning to rank algorithms for information retrieval (IR). We point out that the key for addressing the learning problem is to look at it fr