Prediction of future input explains lateral connectivity in primary visual cortex

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neurons in the primary visual cortex (V1) exhibit specific functional connections.
• Existing research highlights wiring biases related to V1 neuron tuning properties.
• The underlying organizational principles of these connectivity rules remain unclear.

Purpose of the Study:

• To investigate if a unifying principle explains the functional specificity of V1 connections.
• To determine if temporal prediction can naturally generate observed V1 wiring patterns.
• To provide a principled explanation for the functional and anatomical properties of early sensory cortex.

Main Methods:

• Developed a recurrent neural network model.
• Trained the network to predict upcoming sensory inputs using natural visual stimuli.
• Analyzed the emergent connectivity patterns and their relationship to neuronal tuning properties.

Main Results:

• The temporal prediction model successfully reproduced complex relationships between V1 neuron connectivity and orientation/direction preferences.
• The model demonstrated that highly connected neurons tend to respond more similarly to natural movies.
• Differences in functional connectivity between excitatory and inhibitory V1 populations were also replicated.

Conclusions:

• Functional specificity in V1 connections naturally emerges from a temporal prediction objective.
• This principle offers a unified explanation for the observed functional and anatomical organization of the early visual cortex.
• The findings suggest that predictive coding may be a fundamental principle underlying cortical circuit development.