Structures of Neural Correlation and How They Favor Coding

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural information processing is limited by trial-to-trial response variability, known as noise.
• The role of correlated noise in population coding has been theoretically debated but experimentally underexplored.

Purpose of the Study:

• To experimentally validate the impact of correlated noise on neural population coding.
• To refine theoretical models of neural coding by incorporating stimulus-dependent noise correlations.

Main Methods:

• Simultaneous recordings from populations of direction-selective retinal ganglion cells.
• Development of functional models to capture stimulus-dependent noise statistics.
• Quantification of population coding performance based on noise correlations and feature sensitivities.

Main Results:

• Correlated neural noise beneficially impacts population coding, even in small neural populations.
• The stimulus-dependent structure of noise correlations is crucial for this coding benefit.
• Favorable correlation structures emerge robustly in neural circuits with noisy, nonlinear components.

Conclusions:

• Correlated noise is not merely a nuisance but a functional component of neural coding.
• The findings provide a refined theoretical framework for understanding neural population coding.
• The principles observed in the retina are likely applicable to other neural systems.