Locations in the Neocortex: A Theory of Sensorimotor Object Recognition Using Cortical Grid Cells

Area of Science:

• Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• The neocortex anticipates sensory consequences of movement, but neural mechanisms remain unclear.
• Entorhinal cortex grid cells track an animal's location via path integration.
• Understanding how sensory neocortex integrates movement is a key challenge.

Purpose of the Study:

• To propose a neural network model for how the sensory neocortex incorporates movement.
• To explain how grid cell-like neurons can represent object-specific locations.
• To demonstrate robust object learning and sensory prediction using movement cues.

Main Methods:

• A two-layer neural network model was developed using cortical grid cells and path integration principles.
• The model uses object-specific location representations and sensory input processing.
• Simulations were performed to test the model's learning and recognition capabilities.

Main Results:

• The model successfully learned and recognized hundreds of objects, even with ambiguous features.
• It demonstrated robust object recognition by integrating sensory and motor information.
• The model predicts sensory stimuli based on learned object locations and movement.

Conclusions:

• Grid cell-like mechanisms in the neocortex can represent object-specific locations updated by movement.
• This model provides a framework for understanding sensory neocortex function in movement-based learning and prediction.
• The proposed model aligns with known cortical circuitry, particularly reciprocal connections between layers 4 and 6.