Reward prediction errors, not sensory prediction errors, play a major role in model selection in human reinforcement learning.
Yihao WuMasahiko MoritaJun IzawaPublished in: Neural Networks (2022)
Keyphrases
- model selection
- play a major role
- reinforcement learning
- cross validation
- hyperparameters
- machine learning
- sample size
- bayesian learning
- statistical learning
- statistical inference
- model selection criteria
- generalization error
- parameter estimation
- regression model
- optimal policy
- variable selection
- information criterion
- selection criterion
- sensory inputs
- marginal likelihood
- gaussian process
- reward function
- feature selection
- bayesian methods
- meta learning
- motion segmentation
- mixture model
- state space
- markov decision processes
- error estimation
- bayesian information criterion
- supervised learning
- leave one out cross validation
- image registration
- semi supervised
- dynamic programming
- parameter determination