Sparse coding in striate and extrastriate visual cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience

Background:

• Theoretical studies suggest efficient neural codes in the mammalian cortex should be sparse.
• Previous research used varying definitions of
• sparse,
• leading to assumptions about neural coding.

Purpose of the Study:

• To examine and test assumptions about neural sparseness in primate visual cortex.
• To investigate the relationship between lifetime and population sparseness.
• To determine if neural codes are optimized for maximum lifetime sparseness.

Main Methods:

• Measuring lifetime sparseness in macaque visual cortex (V1, V2, V4) during natural image presentation.
• Analyzing action potential counts and firing rates.
• Testing the correlation between lifetime and population sparseness.

Main Results:

• Lifetime and population sparseness are not necessarily correlated.
• High lifetime sparseness can occur irrespective of action potential count.
• Lifetime sparseness did not increase across the visual hierarchy (V1, V2, V4).
• Neural responses are often described by exponential distributions, consistent with metabolic constraints.

Conclusions:

• The neural code may not be optimized solely for maximizing lifetime sparseness.
• Neurons might be optimized for information transmission under metabolic constraints on firing rate.
• Existing assumptions about sparse coding in the mammalian cortex require re-evaluation.