English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

NDCG, namely Normalized Discounted Cumulative Gain, is a widely used rankingmetric in information retrieval and machine learning. However, efficient andprovable stochastic methods for maximizing NDCG are still lacking, especiallyfor deep models. In this paper, we propose a principled approach to optimizeNDCG and its top-$K$ variant. First, we formulate a novel compositionaloptimization problem for optimizing the NDCG surrogate, and a novel bilevelcompositional optimization problem for optimizing the top-$K$ NDCG surrogate.Then, we develop efficient stochastic algorithms with provable convergenceguarantees for the non-convex objectives. Different from existing NDCGoptimization methods, the per-iteration complexity of our algorithms scaleswith the mini-batch size instead of the number of total items. To improve theeffectiveness for deep learning, we further propose practical strategies byusing initial warm-up and stop gradient operator. Experimental results onmultiple datasets demonstrate that our methods outperform prior rankingapproaches in terms of NDCG. To the best of our knowledge, this is the firsttime that stochastic algorithms are proposed to optimize NDCG with a provableconvergence guarantee. Our proposed methods are implemented in the LibAUClibrary at https://libauc.org/.

Large-scale Stochastic Optimization of NDCG Surrogates for Deep Learning with Provable Convergence