Memorize, Factorize, or be Naive: Learning Optimal Feature Interaction Methods for CTR Prediction.
Fuyuan LyuXing TangHuifeng GuoRuiming TangXiuqiang HeRui ZhangXue LiuPublished in: ICDE (2022)
Keyphrases
- learning process
- preprocessing
- learning systems
- significant improvement
- empirical studies
- linear predictors
- prediction accuracy
- learning tasks
- feature subset
- human computer interaction
- selecting features
- feature selection
- learned models
- exhaustive search
- human experts
- machine learning methods
- benchmark datasets
- supervised learning
- dynamic programming
- active learning
- feature vectors
- support vector
- reinforcement learning