Distributional offline continuous-time reinforcement learning with neural physics-informed PDEs (SciPhy RL for DOCTR-L).
Igor HalperinPublished in: Neural Comput. Appl. (2024)
Keyphrases
- reinforcement learning
- state space
- optimal control
- fitted q iteration
- semi markov decision process
- partial differential equations
- rl algorithms
- model free
- network architecture
- reinforcement learning algorithms
- co occurrence
- function approximation
- neural network
- temporal difference
- computer science
- control problems
- reinforcement learning methods
- sensory inputs
- optimal policy
- markov decision process
- partially observable
- numerical solution
- partially observable domains
- machine learning
- anisotropic diffusion
- learning algorithm
- multi agent
- level set
- markov chain
- stochastic processes
- markov decision processes
- dynamic programming
- adaptive control
- image denoising
- markov processes
- state action
- control policy
- function approximators
- policy iteration
- markov decision problems
- actor critic
- hierarchical reinforcement learning
- multiscale
- image processing
- direct policy search
- dynamical systems