From basic network principles to neural architecture: emergence of orientation columns

Area of Science:

• Computational neuroscience
• Neuroscience
• Developmental neuroscience

Background:

• Orientation-selective cells are fundamental to mammalian visual cortex architecture.
• Previous work demonstrated spontaneous emergence of these cells via Hebbian learning in developing networks.
• The role of network connectivity in organizing these cells remained an open question.

Purpose of the Study:

• To investigate the self-organization of orientation-selective cells.
• To determine if lateral connections lead to structured orientation maps.
• To model the formation of orientation columns in the visual cortex.

Main Methods:

• Development of a multilayered neural network model.
• Simulation of Hebbian-type learning rules for feedforward connections.
• Introduction and analysis of lateral connections between developing orientation cells.

Main Results:

• Developing orientation cells self-organized into banded patterns.
• These patterns resemble the orientation columns observed in mammalian visual cortex.
• The organization emerged without orientation preference pre-specification or visual input.

Conclusions:

• Lateral connections are crucial for the self-organization of orientation-selective cells into columns.
• The model demonstrates a plausible mechanism for the emergence of visual cortex architecture.
• These findings align with observations in biological systems, including macaque monkeys.