An Energy-Efficient Deep Reinforcement Learning FPGA Accelerator for Online Fast Adaptation with Selective Mixed-precision Re-training.
Wooyoung JoJuhyoung LeeSeunghyun ParkHoi-Jun YooPublished in: A-SSCC (2021)
Keyphrases
- reinforcement learning
- online learning
- field programmable gate array
- online training
- real time
- batch mode
- wireless sensor networks
- test set
- function approximation
- supervised learning
- training samples
- learning algorithm
- high speed
- dynamic programming
- precision and recall
- hardware implementation
- training process
- training set
- adaptation process
- covariate shift
- balancing exploration and exploitation
- signal processing
- sensor networks
- parallel implementation
- parallel computing
- learning process
- online algorithms
- e learning