Hippocampus, cortex, and basal ganglia: insights from computational models of complementary learning systems

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• The brain's cognitive and behavioral functions arise from complex interactions between distinct neural structures.
• Understanding these interactions requires integrating computational principles with empirical neuroscience data.

Purpose of the Study:

• To present a computational framework explaining the collaborative roles of the hippocampus, neocortex, and basal ganglia in mammalian cognition.
• To elucidate how computational tradeoffs in neural networks underpin different learning functions within these brain regions.

Main Methods:

• Development of a computational framework based on neural network models.
• Analysis of computational tradeoffs related to activity levels, learning rates, and pattern overlap.
• Implementation of the framework in functioning computational models.

Main Results:

• Dissociation of the hippocampus from the cortex based on activity levels, learning rates, and pattern overlap.
• Identification of specialized neural functions within the frontal cortex and basal ganglia system.
• Successful explanation of diverse neuroscience data including recognition memory, fear conditioning, and flexible responding.

Conclusions:

• The proposed cognitive architecture, grounded in computational tradeoffs, provides a robust model for understanding brain function.
• This framework offers a nuanced explanation for a wide range of behavioral and cognitive neuroscience findings.
• The model highlights the specialized yet integrated roles of the hippocampus, neocortex, and basal ganglia in learning and behavior.