Small-correlation expansion to quantify information in noisy sensory systems

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Information Theory

Background:

• Neural networks encode stimuli through collective spiking activity.
• Population responses exhibit noise and complex correlations, hindering information transmission analysis.
• Existing methods are limited to small neuronal groups due to dimensionality challenges.

Purpose of the Study:

• To develop a scalable method for computing stimulus information in large neural populations.
• To analytically characterize the influence of neuronal correlations on information encoding.
• To apply the method to real-world neural data for understanding noise and memory effects.

Main Methods:

• Developed a small-correlation expansion approximation.
• Derived analytical expressions for stimulus information based on firing rates and pairwise correlations.
• Validated the approximation using synthetic data and electrophysiological recordings from the vertebrate retina.

Main Results:

• The small-correlation expansion accurately computes stimulus information in large neural populations.
• The method provides interpretable analytical expressions.
• Quantified the impact of noise correlations and single-neuron memory on information transmission in retinal data.

Conclusions:

• The developed approximation overcomes the curse of dimensionality for analyzing information in large neural populations.
• This approach offers a powerful tool for dissecting the roles of correlations and single-neuron properties in neural coding.
• The findings have implications for understanding neural computation in sensory systems like the retina.