Task-Related Synaptic Changes Localized to Small Neuronal Population in Recurrent Neural Network Cortical Models

Area of Science:

• Computational neuroscience
• Cognitive neuroscience
• Machine learning

Background:

• Humans exhibit flexible cognitive control, with the prefrontal cortex (PFC) implicated.
• Detailed neural mechanisms underlying cognitive flexibility remain largely unknown.
• Recurrent neural network models offer insights into PFC function.

Purpose of the Study:

• To identify critical neural network features for cognitive flexibility.
• To compare four distinct PFC models on a context-dependent integration task.

Main Methods:

• Utilized recurrent neural network models to simulate PFC function.
• Trained four different models on a behavioral task requiring context-dependent integration.
• Analyzed synaptic plasticity and neuronal unit contributions to task performance.

Main Results:

• All models exhibited highly plastic synapses in specific neuronal populations.
• Neuronal units with plastic synapses were crucial for performance in most models.
• No consistent patterns in synaptic strength distribution were found across models.

Conclusions:

• Task-dependent synaptic plasticity is vital for flexible cognition.
• The dynamic changes in synapses are more critical than static network architecture.
• Computational modeling aids in understanding unobservable neural parameters.