PoPS: Policy Pruning and Shrinking for Deep Reinforcement Learning.
Dor LivneKobi CohenPublished in: IEEE J. Sel. Top. Signal Process. (2020)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- reinforcement learning problems
- partially observable
- action selection
- actor critic
- partially observable environments
- function approximators
- function approximation
- approximate dynamic programming
- control policy
- policy evaluation
- reinforcement learning algorithms
- policy gradient
- rl algorithms
- search space
- markov decision processes
- state action
- action space
- policy iteration
- reward function
- multi agent
- state space
- partially observable markov decision processes
- dynamic programming
- long run
- pruning method
- state and action spaces
- model free reinforcement learning
- average reward
- learning algorithm
- optimal control
- reinforcement learning methods
- markov decision problems
- control problems
- model free
- learning problems
- decision problems
- state dependent
- temporal difference
- control policies
- continuous state spaces
- partially observable domains
- machine learning
- policy gradient methods
- agent receives
- neural network
- effective pruning
- agent learns
- inverse reinforcement learning
- transition model
- policy making