Tuning curve sharpening for orientation selectivity: coding efficiency and the impact of correlations

Area of Science:

• Computational neuroscience
• Neural coding and information theory

Background:

• Bell-shaped tuning curves are fundamental to neural population coding.
• Decreasing tuning curve width is often assumed to enhance information transmission.
• This assumption relies on specific, potentially invalid, noise independence conditions.

Purpose of the Study:

• To reexamine the impact of tuning curve sharpening on population codes in spiking neural networks.
• To investigate the role of cortical lateral interactions in tuning curve sharpening and its consequences.
• To compare models with and without sharpening via lateral interactions using orientation selectivity as a case study.

Main Methods:

• Simulations of spiking neural networks with and without sharpening via lateral interactions.
• Analysis of orientation selectivity to assess population coding efficiency.
• Evaluation of information content and code complexity under different sharpening conditions.

Main Results:

• Sharpening tuning curves through lateral interactions does not improve, but significantly degrades, population codes.
• Sharpening models introduce complex codes heavily dependent on pairwise neuronal correlations.
• Information loss is a critical consequence of sharpening via lateral interactions.

Conclusions:

• The common notion that sharpening tuning curves improves population codes is challenged.
• Cortical lateral interactions, when sharpening tuning curves, impair neural information processing.
• The study provides experimental predictions to differentiate between sharpening and non-sharpening network models.