RAPID-RL: A Reconfigurable Architecture with Preemptive-Exits for Efficient Deep-Reinforcement Learning.
Adarsh Kumar KostaMalik Aqeel AnwarPriyadarshini PandaArijit RaychowdhuryKaushik RoyPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- reconfigurable architecture
- function approximation
- state space
- optimal policy
- markov decision processes
- learning process
- transfer learning
- lower bound
- model free
- systolic array
- learning problems
- rl algorithms
- learning algorithm
- action selection
- partially observable
- markov decision process
- function approximators
- continuous state
- image segmentation