Pseudosparse neural coding in the visual system of primates.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• The efficient coding hypothesis posits that neural population responses to stimuli are sparse.
• Sparse coding assumes that different stimuli activate distinct, uncorrelated sets of neurons.
• This assumption is crucial for interpreting neural population activity as efficient.

Purpose of the Study:

• To investigate the degree of correlation in neural population responses across various brain regions in macaque monkeys.
• To introduce and evaluate the 'pseudosparseness index' as a measure of population response correlation.
• To determine if neural population activity exhibits authentic sparseness or high pseudosparseness.

Main Methods:

• Analysis of neurophysiological data from multiple macaque cortical areas (V1, V2, MT, aIT, LIP, FEF, PRh).
• Calculation of the pseudosparseness index, defined as the mean correlation of population responses.
• Comparison of pseudosparseness values with authentic sparseness (0.00) across different stimulus types and recording methods.

Main Results:

• Consistently high pseudosparseness values (0.59-0.98) were observed across all examined brain regions and datasets.
• Pseudosparseness levels were significantly higher than authentic sparseness across synthetic and natural stimuli.
• A computational model identified the standard deviation of spontaneous neural activity as a key factor driving high pseudosparseness.

Conclusions:

• Neural population responses in the macaque cortex exhibit high pseudosparseness, contradicting the assumption of uncorrelated responses.
• The prevalence of pseudosparseness challenges the widespread application of sparse coding frameworks in understanding neural computation.
• Re-evaluation of the sparse coding hypothesis and its applicability to cortical neural coding is warranted.