Learning-to-Rank at the Speed of Sampling: Plackett-Luce Gradient Estimation with Minimal Computational Complexity.
Harrie OosterhuisPublished in: SIGIR (2022)
Keyphrases
- learning to rank
- gradient estimation
- computational complexity
- variance reduction
- active sampling
- loss function
- ranking functions
- information retrieval
- ranking svm
- document retrieval
- collaborative filtering
- direct optimization
- sample size
- evaluation measures
- monte carlo
- ranking models
- decision problems
- test collection
- learning to rank algorithms
- supervised learning
- np hard
- active learning
- pairwise
- decision trees