Hybrid Reward Architecture for Reinforcement Learning.
Harm van SeijenMehdi FatemiJoshua RomoffRomain LarocheTavian BarnesJeffrey TsangPublished in: CoRR (2017)
Keyphrases
- reinforcement learning
- learning capabilities
- function approximation
- state space
- machine learning
- management system
- transfer learning
- model free
- temporal difference
- reward shaping
- markov decision processes
- neural network
- policy search
- average reward
- architectural design
- learning agent
- reinforcement learning algorithms
- reward function
- software architecture
- dynamic programming
- real time
- real robot
- hybrid learning
- learning algorithm