Gaussian Processes for Fast Policy Optimisation of POMDP-based Dialogue Managers.
Milica GasicFilip JurcícekSimon KeizerFrançois MairesseBlaise ThomsonKai YuSteve J. YoungPublished in: SIGDIAL Conference (2010)
Keyphrases
- gaussian processes
- optimal policy
- model free reinforcement learning
- partially observable markov decision processes
- partially observable
- partially observable markov decision process
- markov decision process
- gaussian process
- reinforcement learning
- gaussian process regression
- covariance function
- state space
- reward function
- decision problems
- markov decision processes
- preference learning
- infinite horizon
- belief state
- gaussian process models
- multi class
- dynamic programming
- decision trees
- multi task
- learning tasks
- e learning