Reinforcement Learning for Slate-based Recommender Systems: A Tractable Decomposition and Practical Methodology.
Eugene IeVihan JainJing WangSanmit NarvekarRitesh AgarwalRui WuHeng-Tze ChengMorgane LustmanVince GattoPaul CovingtonJim McFaddenTushar ChandraCraig BoutilierPublished in: CoRR (2019)
Keyphrases
- recommender systems
- reinforcement learning
- collaborative filtering
- computational complexity
- function approximation
- real world
- user profiling
- markov decision processes
- user preferences
- machine learning
- hypertree decomposition
- reinforcement learning algorithms
- transfer learning
- learning algorithm
- information retrieval
- user profiles
- np complete
- relevance feedback
- website
- information filtering
- user modeling
- temporal difference
- knowledge base
- decomposition method
- cold start problem