An incremental Hebbian learning model of the primary visual cortex with lateral plasticity and real input patterns

Area of Science:

• Computational neuroscience
• Neuroscience and neurobiology
• Artificial intelligence

Background:

• The primary visual cortex (V1) exhibits complex receptive fields and organized maps crucial for visual processing.
• Understanding the self-organization principles underlying V1 map development is a key challenge in neuroscience.
• Neural network models offer a powerful framework to investigate these emergent properties.

Purpose of the Study:

• To develop and analyze a simplified binocular neural network model of the primary visual cortex.
• To investigate the role of Hebbian learning rules and synaptic plasticity in map formation.
• To explore how different stimulus patterns influence the development of orientation and ocular dominance maps.

Main Methods:

• A simplified binocular neural network model with separate ON/OFF pathways was implemented.
• Hebbian learning rules, stabilized by constant norm and sum constraints, governed synaptic weight adaptation.
• Simulations utilized both random and natural grayscale image stimuli to drive network development.

Main Results:

• Realistic orientation maps, including +/- 1/2-vortices, developed under random input patterns.
• Plastic lateral couplings self-organized into Mexican hat-like structures on average.
• Natural image inputs led to realistic orientation maps, with lateral couplings reflecting input image correlations.

Conclusions:

• The model successfully reproduces key features of primary visual cortex organization, such as orientation maps.
• Hebbian learning, constrained by normalization, is sufficient for self-organizing cortical maps.
• The model highlights the relationship between statistical properties of natural images and emergent neural structures.