Distributional Offline Continuous-Time Reinforcement Learning with Neural Physics-Informed PDEs (SciPhy RL for DOCTR-L).
Igor HalperinPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- state space
- fitted q iteration
- optimal control
- partial differential equations
- semi markov decision process
- function approximation
- network architecture
- rl algorithms
- reinforcement learning algorithms
- markov chain
- neural network
- computer science
- machine learning
- sensory inputs
- level set
- multi agent
- co occurrence
- model free
- markov decision processes
- temporal difference
- dynamic programming
- temporal difference learning
- image denoising
- image processing
- action selection
- control problems
- markov processes
- learning algorithm
- actor critic
- function approximators
- numerical solution
- adaptive control
- optimal policy
- stochastic processes
- learning classifier systems
- complex domains
- markov decision process
- continuous state
- direct policy search
- partially observable domains
- reward shaping
- hierarchical reinforcement learning
- policy search
- image segmentation
- state action
- average reward
- partially observable markov decision processes