Direct Learning to Rank and Rerank
This addresses a bottleneck in prioritization problems for applications like search and recommendation, but appears incremental as it builds on existing learning-to-rank frameworks.
The paper tackles the problem of poor approximations in learning-to-rank algorithms due to convex proxies, proposing 'exact' reranking methods based on mathematical programming and proving a relaxed version yields the same optimal solution, with empirical analysis provided.
Learning-to-rank techniques have proven to be extremely useful for prioritization problems, where we rank items in order of their estimated probabilities, and dedicate our limited resources to the top-ranked items. This work exposes a serious problem with the state of learning-to-rank algorithms, which is that they are based on convex proxies that lead to poor approximations. We then discuss the possibility of "exact" reranking algorithms based on mathematical programming. We prove that a relaxed version of the "exact" problem has the same optimal solution, and provide an empirical analysis.