Intrinsic Rewards for Exploration without Harm from Observational Noise: A Simulation Study Based on the Free Energy Principle.
Theodore Jerome TinkerKenji DoyaJun TaniPublished in: CoRR (2024)
Keyphrases
- simulation study
- free energy
- belief propagation
- fixed point
- upper bound
- monte carlo
- posterior distribution
- competitive learning
- graphical models
- approximate inference
- reinforcement learning
- missing data
- noise level
- noise reduction
- energy minimization
- similarity measure
- probabilistic model
- lower bound
- prior knowledge
- closed form
- machine learning
- sufficient conditions
- graph cuts
- maximum likelihood