Modelling transition dynamics in MDPs with RKHS embeddings.
Steffen GrünewälderGuy LeverLuca BaldassarreMassimiliano PontilArthur GrettonPublished in: ICML (2012)
Keyphrases
- markov decision processes
- reproducing kernel hilbert space
- transition model
- reinforcement learning
- kernel methods
- state space
- euclidean space
- loss function
- random projections
- vector space
- optimal policy
- decision theoretic planning
- low dimensional
- hilbert space
- markov decision process
- reproducing kernel
- factored mdps
- markov decision problems
- finite horizon
- distance measure
- state transition
- state transitions
- reward function
- dimensionality reduction
- gaussian process
- manifold learning