Identify stimulus ensembles that benefit most from noise correlations

Determine which stimulus ensembles most benefit from noise correlations in neural population coding, and establish where naturalistic stimuli stand relative to those ensembles to clarify the conditions under which noise correlations enhance information transmission.

Background

The paper develops a theoretical and experimental framework showing that strong noise correlations can enhance information transmission in neural populations, even when they share the same sign as signal correlations. The authors validate these predictions in retinal recordings and extend the analysis to large populations, demonstrating that noise correlations preferentially improve the encoding of fine-scale stimulus features while sometimes degrading large-scale features.

Given these findings, a central unresolved issue is to delineate which classes of stimuli (e.g., different spatial or temporal statistics, synthetic versus naturalistic) yield maximal benefit from noise correlations, and to position naturalistic stimuli within this landscape. Addressing this question is crucial for understanding how correlated variability aids real-world sensory processing.