Nonlinear stimulus representations in neural circuits with approximate excitatory-inhibitory balance

Area of Science:

• Computational neuroscience
• Neural dynamics
• Systems neuroscience

Background:

• Cortical circuits exhibit a balance between excitation and inhibition.
• Dynamically balanced network models predict linear stimulus-response relationships.
• The mechanism for nonlinear computations in the cortex remains unclear.

Purpose of the Study:

• To investigate how cortical circuits implement nonlinear representations and computations.
• To explore the consequences of deviations from the balanced state in neuronal networks.

Main Methods:

• Analysis of computational models of neuronal networks.
• Mathematical characterization of network states under varying stimuli.
• Comparison of model predictions with experimental cortical recordings.

Main Results:

• Every balanced network architecture can be driven into a "semi-balanced state" by specific stimuli.
• This semi-balanced state is characterized by excess inhibition and absence of excess excitation.
• The semi-balanced state generates nonlinear stimulus representations and computations, is unavoidable with multiple stimuli, and aligns with cortical data.

Conclusions:

• Deviations from the balanced state, leading to a semi-balanced state, are crucial for nonlinear neural processing.
• The semi-balanced state provides a mathematical framework explaining nonlinear computations in the brain and artificial neural networks.