Optimistic Simulated Exploration as an Incentive for Real Exploration

Ivo Danihelka

Optimistic Simulated Exploration as an Incentive for Real Exploration

Machine Learning 2009-05-20 v3 Artificial Intelligence

Authors: Ivo Danihelka

Abstract

Many reinforcement learning exploration techniques are overly optimistic and try to explore every state. Such exploration is impossible in environments with the unlimited number of states. I propose to use simulated exploration with an optimistic model to discover promising paths for real exploration. This reduces the needs for the real exploration.

Keywords

reinforcement learning

Cite

@article{arxiv.0903.2972,
  title  = {Optimistic Simulated Exploration as an Incentive for Real Exploration},
  author = {Ivo Danihelka},
  journal= {arXiv preprint arXiv:0903.2972},
  year   = {2009}
}

Comments

accepted, noted that the initial path was 217 steps long

Related papers

View all related →

Machine Learning · Statistics

On Optimistic versus Randomized Exploration in Reinforcement Learning