A role for working memory in shaping the action policy for reinforcement learning.
Ham HuangSamuel McdougleAnne CollinsPublished in: CogSci (2020)
Keyphrases
- working memory
- reinforcement learning
- reward shaping
- action selection
- cognitive load
- optimal policy
- action space
- markov decision problems
- agent learns
- state action
- computational model
- information processing
- focus of attention
- partially observable domains
- policy search
- basal ganglia
- prefrontal cortex
- long term memory
- reinforcement learning algorithms
- state space
- cognitive architecture
- markov decision process
- short term memory
- reward function
- partially observable
- individual differences
- function approximators
- markov decision processes
- functional connectivity
- working memory capacity
- policy gradient
- learning algorithm
- reward signal
- agent receives
- partially observable markov decision processes
- real time
- decision making
- machine learning