A study by researchers at the Korea Advanced Institute of Science and Technology (KAIST) in South Korea and the California Institute of Technology helps explain how the human brain adapts to complexity in learning and decision-making.
The researchers uncovered a computational and neural mechanism for human meta-reinforcement learning. They applied reinforcement learning theory-based experiment design to optimize three factors—goal, task complexity, and task uncertainty—of the two-stage Markov decision task, manipulating confounding variables and modeling the situation after actual human problem-solving.
The researchers also used model-based neuroimaging analysis and parameter recovery analysis to yield a meta-reinforcement learning computational model.
KAIST's Sang Wan Lee said, "This study ... holds significant potential for applying core insights gleaned into how human intelligence works with algorithms."
From Korea Advanced Institute of Science and Technology
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found