Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference.
Ranggi HwangJianyu WeiShijie CaoChangho HwangXiaohu TangTing CaoMao YangPublished in: ISCA (2024)
Keyphrases
- preprocessing
- dynamic programming
- detection algorithm
- learning algorithm
- times faster
- expectation maximization
- high accuracy
- worst case
- computational cost
- k means
- improved algorithm
- cost function
- experimental evaluation
- significant improvement
- np hard
- input data
- optimal solution
- theoretical analysis
- matching algorithm
- single pass
- recognition algorithm
- convergence rate
- bayesian framework
- data sets
- ant colony optimization
- reinforcement learning
- optimization algorithm
- em algorithm
- computational complexity