Improving Generalization in Meta Reinforcement Learning using Learned Objectives.
Louis KirschSjoerd van SteenkisteJürgen SchmidhuberPublished in: CoRR (2019)
Keyphrases
- data sets
- reinforcement learning
- efficient learning
- training data
- function approximation
- stochastic approximation
- learning capabilities
- state space
- transfer learning
- robot control
- reinforcement learning algorithms
- learning classifier systems
- multiple objectives
- markov decision processes
- optimal policy
- probabilistic model
- information systems
- learning algorithm
- neural network