Topographic receptive fields and patterned lateral interaction in a self-organizing model of the primary visual...

Area of Science:

• Computational neuroscience
• Artificial neural networks
• Cortical map development

Background:

• Cortical maps exhibit topographic organization and complex lateral interactions.
• The developmental mechanisms underlying these structures remain incompletely understood.
• Existing models often struggle to integrate receptive field and lateral connection development.

Purpose of the Study:

• To present a novel self-organizing neural network model.
• To investigate the simultaneous and cooperative development of topographic receptive fields and lateral interactions.
• To elucidate the role of local, unsupervised learning in cortical map formation.

Main Methods:

• A self-organizing neural network model was developed.
• Both afferent and lateral connections were adapted using a Hebbian learning mechanism.
• The learning process was purely local and unsupervised.
• Receptive field and lateral interaction profiles were analyzed.

Main Results:

• Neurons self-organized into topographic receptive fields with hill-shaped input weight profiles.
• Unique and patterned lateral interaction profiles emerged for each neuron.
• The model demonstrated that lateral connections develop based on correlated neural activity.
• Developed lateral connection patterns were shown to correlate with receptive field properties like ocular dominance.

Conclusions:

• A unified, local learning mechanism can account for the cooperative development of receptive fields and lateral connections.
• The model provides a potential explanation for the emergence of specific connection patterns in cortical maps.
• This work offers insights into the self-organization principles governing neural map development.