RL-Scope: Cross-stack Profiling for Deep Reinforcement Learning Workloads.
James GleesonMoshe GabelGennady PekhimenkoEyal de LaraSrivatsan KrishnanVijay Janapa ReddiPublished in: MLSys (2021)
Keyphrases
- reinforcement learning
- function approximation
- state space
- rl algorithms
- reinforcement learning algorithms
- model free
- control problems
- database systems
- markov decision processes
- computer systems
- learning algorithm
- optimal policy
- optimal control
- learning agents
- learning classifier systems
- policy search
- continuous state and action spaces
- temporal difference
- learning process
- direct policy search
- multi agent
- policy iteration
- function approximators
- temporal difference learning
- partially observable
- control policy
- control strategies
- dynamic programming
- access patterns
- transfer learning