Linear NDCG and Pair-wise Loss

TL;DR

This paper demonstrates that the error in linear NDCG, a metric for web content quality assessment, is equivalent to a newly proposed pair-wise loss function.

Contribution

It introduces a novel theoretical connection between linear NDCG and pair-wise loss, providing insights into their relationship.

Findings

Proves the equivalence between DCG error and a new pair-wise loss.

Provides a theoretical foundation for optimizing NDCG using pair-wise loss.

Enhances understanding of evaluation metrics in web content quality assessment.

Abstract

Linear NDCG is used for measuring the performance of the Web content quality assessment in ECML/PKDD Discovery Challenge 2010. In this paper, we will prove that the DCG error equals a new pair-wise loss.