Reward Shaping with Recurrent Neural Networks for Speeding up On-Line Policy Learning in Spoken Dialogue Systems.
Pei-hao SuDavid VandykeMilica GasicNikola MrksicTsung-Hsien WenSteve J. YoungPublished in: CoRR (2015)
Keyphrases
- recurrent neural networks
- learning process
- recurrent networks
- spoken dialogue systems
- learning algorithm
- reinforcement learning
- neural network
- knowledge acquisition
- reward shaping
- echo state networks
- policy iteration
- solving problems
- feed forward
- markov decision processes
- domain knowledge
- hidden markov models
- prior knowledge