A Model of Motion Processing in the Visual Cortex Using Neural Field With Asymmetric Hebbian Learning

Area of Science:

• Computational Neuroscience
• Visual System Modeling

Background:

• Neurons in the dorsal visual pathway process motion, starting in V1 and progressing to MT and MST.
• Understanding hierarchical motion processing is crucial for visual neuroscience.

Purpose of the Study:

• To present a hierarchical neural field network model for visual motion processing.
• To investigate the model's ability to replicate neuronal responses in V1 and MT.
• To demonstrate the model's capacity for processing complex motion stimuli.

Main Methods:

• Developed a hierarchical neural field network with one or two layers (NF1 for V1, NF2 for MT).
• Employed asymmetric Hebbian learning for lateral connections to process sequential motion information.
• Trained the model on diverse stimuli: bars, squares, gratings, plaids, and random dot stimuli (RDS).

Main Results:

• A single-layer network (NF1) developed direction maps for bar stimuli.
• Two-layer networks showed NF1 responding to component motion and NF2 to pattern motion (squares, plaids).
• NF2 successfully encoded translational flow motion from RDS, achieving 100% accuracy on training and 90% on testing data.

Conclusions:

• The proposed neural field network model effectively simulates hierarchical motion processing in the visual cortex.
• The model's ability to generalize and its response properties align with electrophysiological findings.
• Asymmetric Hebbian learning enables the model to capture sequential motion processing capabilities.