网络搜索-填补信息断层.pptVIP

We find that top-k is sufficient for ranking! Motivation: empirical study CIKM’13:Is top-k Sufficient for Ranking? We find that top-k is sufficient for ranking! Loss function in top-k setting is a upper bound of evaluation measure based error Loss function in top-k setting is a lower bound of loss function in full-order setting Motivation: theoretical study CIKM’13:Is top-k Sufficient for Ranking? Top-k Learning to Rank Framework Three Tasks: How to design an efficient labeling strategy to get top-k ground-truth? How to develop more powerful ranking algorithms in the new scenario? How to d