Feedback-induced gain control in stochastic spiking networks.

Area of Science:

• Computational Neuroscience
• Neural Networks
• Systems Neuroscience

Background:

• Gain control mechanisms in neural systems are crucial for processing diverse inputs.
• Experimental systems exhibit mixtures of divisive and subtractive gain control.
• Understanding the origins of these gain control types is essential.

Purpose of the Study:

• Investigate the joint influence of recurrent feedback and noise on gain control.
• Differentiate between divisive and subtractive gain control mechanisms.
• Explore feedback originating from neuronal output spikes.

Main Methods:

• Theoretical analysis and numerical simulations of globally coupled spiking leaky integrate-and-fire neurons.
• Modeling feedback spikes as alpha functions with additive current.
• Analyzing changes in the firing frequency-input bias (f-I) curve slope.

Main Results:

• Additive negative or positive feedback naturally produces divisive gain control.
• Negative feedback lowers gain, while positive feedback can increase gain or cause abrupt jumps.
• Noise alone creates mixed divisive and subtractive gain control.
• Combined feedback and noise primarily result in divisive gain control, demonstrating robustness.

Conclusions:

• Recurrent feedback is a primary driver of divisive gain control in neural networks.
• Gain control mechanisms are robust to the presence of noise.
• Findings align with experimental observations in systems like weakly electric fish.