A general approximation framework for direct optimization of information retrieval measures

By Tao Qin, Tie-Yan Liu, Hang Li. 2010

273 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems, Information Retrieval & Search, Optimization & Training

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Tie-Yan Liu
2. Tao Qin
3. Jiafeng Guo
4. Xueqi Cheng
5. Jun Xu
6. Hang Li
7. Yanyan Lan
8. Jianfeng Gao

In Computer Vision and Pattern Recognition

1. Steven C. H. Hoi
2. Tie-Yan Liu
3. Hang Li
4. Tao Qin
5. Jiafeng Guo
6. Xueqi Cheng
7. Dacheng Tao
8. Jun Xu

In Information Systems

1. Hang Li
2. Jun Xu
3. Tie-Yan Liu
4. Tao Qin
5. Yanyan Lan
6. Xueqi Cheng
7. Michael R. Lyu
8. Jiafeng Guo

In Information Retrieval & Search

1. Hang Li
2. Yanyan Lan
3. Tie-Yan Liu
4. Xueqi Cheng
5. Jun Xu
6. Liang Pang
7. Tao Qin
8. Jiafeng Guo

In Optimization & Training

1. Hang Li
2. Tie-Yan Liu
3. Tao Qin
4. Irwin King
5. Rohan Anil
6. Jingtao Zhan
7. Jiaxin Mao
8. Xueqi Cheng

Citations in the corpus

• VD-BERT: A Unified Vision and Dialog Transformer with BERT (2020, methodology): “In Table 3c, we compare Cross Entropy (CE) training with several other listwise ranking methods: ListNet (Cao et al., 2007), ListMLE (Xia et al., 2008), and approx-NDCG (Qin et al., 2010).”
• Optimizing Dense Retrieval Model Training with Hard Negatives (2021, methodology): “[25], when the training set is large enough, this method will yield the optimal model in terms of the corresponding metric.”
• TF-Ranking: Scalable TensorFlow Library for Learning-to-Rank (2018, background): “They do so by, among other techniques, creating a smooth variant of popular ranking metrics [35, 39]; deriving tight upper-bounds on ranking metrics [42]; bypassing the requirement that a loss function be defined altogether [6]; or, otherwise designing losses that are loosely related to ranking metr”
• Fast Differentiable Sorting and Ranking (2020, methodology): “• All-pairs (Qin et al., 2010): observing that [ r ( θ )] i can be written as one can obtain soft ranks in O ( n 2 ) by replacing the indicator function with a sigmoid.”
• Learning to Rank for Information Retrieval and Natural Language Processing (2011, unknown): “PermuRank [110] ListMLE [104] AppRank [81]”
• Direct Learning to Rank and Rerank (2018, unknown): “This thought has been considered in several previous works on learning-to-rank (Le et al, 2010; Ertekin and Rudin, 2011; Tan et al, 2013; Chakrabarti et al, 2008; Qin et al, 2013).”
• Learning to rank for information retrieval (2009, unknown): “Example algorithms include SoftRank [119] and AppRank [98].”
• LETOR: A benchmark collection for research on learning to rank for information retrieval (2010, methodology): “Representative work includes ListNet (Cao et al. 2007), Rank- Cosine (Qin et al. 2008e), relational ranking (Qin et al. 2008d), global ranking (Qin et al. 2008c), and StructRank (Huang et al. 2009).”
• L2R²: Leveraging Ranking for Abductive Reasoning (2020, methodology): “al cases of pairwise ranking methods when N = 2. Listwise loss functions are defined on the basis of lists of hypotheses. In this paper, the loss functions of ListNet [4], ListMLE [11] and ApproxNDCG [10] are employed. In the ListNet approach, K-L divergence between the permutation probability distr”
• A Survey of Model Architectures in Information Retrieval (2025, unknown): “Recent works such as GSF (Ai et al., 2019) and A PPROX NDCG (Bruch et al., 2019) use multiple fully connected layers.”