Are single-peaked tuning curves tuned for speed rather than accuracy?

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Information theory

Background:

• The efficient coding hypothesis posits sensory neurons maximize environmental information under biophysical constraints.
• Early visual areas typically show single-peaked neural tuning curves.
• Periodic tuning, seen in grid cells, enhances decoding performance, raising questions about visual area optimality.

Purpose of the Study:

• To investigate the role of temporal coding scales in understanding single-peaked versus periodic tuning curves.
• To explore how decoding time and stimulus dimensionality influence optimal tuning curve shapes for error reduction.
• To provide normative explanations for the prevalence of single-peaked tuning in early visual areas.

Main Methods:

• Analysis of the trade-off between decoding time and decoding ability introduced by catastrophic errors.
• Investigation of tuning curve spatial periods, specifically for circular tuning curves.
• Mathematical modeling to assess the impact of stimulus dimensionality and ongoing neural activity.

Main Results:

• A trade-off exists between decoding speed and accuracy due to the potential for large errors.
• Minimal decoding time generally increases with higher Fisher information, indicating an accuracy-speed trade-off.
• This trade-off is exacerbated by high stimulus dimensionality and ongoing neural activity.

Conclusions:

• The observed single-peaked tuning in early visual areas can be normatively explained by constraints on processing speed.
• Efficient neural coding involves balancing information transmission rate with the minimization of significant errors.
• Understanding the temporal dynamics of neural encoding is crucial for deciphering the functional advantages of different tuning structures.