Rethinking Ranking-based Loss Functions: Only Penalizing Negative Instances before Positive Ones is Enough.
Zhuo LiWeiqing MinJiajun SongYaohui ZhuShuqiang JiangPublished in: CoRR (2021)
Keyphrases
- loss function
- learning to rank
- positive and negative
- learning to rank algorithms
- pairwise
- support vector
- ranking svm
- squared error
- ranking models
- logistic regression
- risk minimization
- web search
- ranking algorithm
- ranking functions
- loss minimization
- evaluation measures
- hinge loss
- reproducing kernel hilbert space
- learning algorithm
- convex loss functions
- boosting algorithms
- pairwise constraints
- ordinal regression
- positive examples
- evaluation metrics
- document retrieval
- information retrieval