NENYA: Cascade Reinforcement Learning for Cost-Aware Failure Mitigation at Microsoft 365.
Lu WangPu ZhaoChao DuChuan LuoMengna SuFangkai YangYudong LiuQingwei LinMin WangYingnong DangHongyu ZhangSaravan RajmohanDongmei ZhangPublished in: KDD (2022)
Keyphrases
- reinforcement learning
- failure rate
- function approximation
- high cost
- total cost
- machine learning
- learning algorithm
- expected cost
- data sets
- database applications
- transfer learning
- model free
- cost reduction
- reinforcement learning algorithms
- learning classifier systems
- optimal control
- risk management
- learning problems
- markov decision processes
- optimal policy
- neural network
- databases