Value Penalized Q-Learning for Recommender Systems.
Chengqian GaoKe XuKuangqi ZhouLanqing LiXueqian WangBo YuanPeilin ZhaoPublished in: SIGIR (2022)
Keyphrases
- cold start problem
- recommender systems
- cold start
- collaborative filtering
- reinforcement learning
- cooperative
- matrix factorization
- state space
- multi agent
- least squares
- item based collaborative filtering
- function approximation
- data sparsity
- model free
- action selection
- learning algorithm
- maximum likelihood
- user preferences
- reinforcement learning algorithms
- optimal policy
- trust aware
- stochastic approximation
- temporal difference learning
- recommendation quality
- user ratings
- information overload
- user modeling
- loss function
- dynamic programming
- model selection
- user profiles
- user profiling
- variable selection
- multi agent reinforcement learning
- potential field
- user model
- compound critiques
- reinforcement learning methods
- user feedback
- user interests
- information filtering
- policy iteration
- implicit feedback
- recommendation algorithms