Soft BPR Loss for Dynamic Hard Negative Sampling in Recommender Systems.
Kexin ShiYun ZhangBingyi JingWenjia WangPublished in: CoRR (2022)
Keyphrases
- recommender systems
- positive and negative
- collaborative filtering
- dynamic environments
- data sets
- user modelling
- random sampling
- monte carlo
- user preferences
- user profiles
- sample size
- decision trees
- learning algorithm
- parameter space
- genetic algorithm
- information overload
- sampling methods
- user profiling
- cold start problem
- database