Reinforcement Learning in Sparse-Reward Environments With Hindsight Policy Gradients.
Paulo E. RauberAvinash UmmadisinguFilipe MutzJürgen SchmidhuberPublished in: Neural Comput. (2021)
Keyphrases
- reinforcement learning
- action selection
- optimal policy
- reward function
- partially observable environments
- policy gradient
- total reward
- policy search
- control policy
- reinforcement learning algorithms
- markov decision process
- average reward
- state action
- state space
- function approximation
- actor critic
- markov decision processes
- partially observable
- eligibility traces
- agent learns
- partially observable markov decision processes
- inverse reinforcement learning
- function approximators
- reinforcement learning problems
- rl algorithms
- policy iteration
- approximate dynamic programming
- control policies
- model free
- long run
- policy evaluation
- multi agent environments
- markov decision problems
- sparse data
- action space
- finite horizon
- discounted reward
- sparse representation
- learning process
- high dimensional
- machine learning
- temporal difference
- finite state
- decision problems
- dynamic environments
- transition model
- agent receives
- partially observable domains
- neural network
- gradient information
- compressive sensing
- autonomous robots
- optimal control
- robotic systems
- dynamic programming
- multi agent
- real world