Rate-distortion theory of neural coding and its implications for working memory

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Information Theory

Background:

• Rate-distortion theory offers a computational framework for memory, defining the trade-off between information rate and distortion (errors).
• Neural population coding models aim to explain how neural activity represents information, but have not fully captured memory regularities.

Purpose of the Study:

• To demonstrate a neural population coding model that realizes the rate-distortion theory framework for memory.
• To explain key regularities in visual working memory using this model.
• To make and test novel predictions about neural activity in memory tasks.

Main Methods:

• Developed a computational model of neural population coding based on rate-distortion principles.
• Applied the model to explain existing data on visual working memory.
• Reanalyzed electrophysiological recordings from monkey prefrontal cortex during a delayed response task.

Main Results:

• The population coding model successfully reproduced known regularities of visual working memory.
• The model explained phenomena not previously accounted for by other population coding models.
• A novel prediction of the model was confirmed by reanalyzing neural data.

Conclusions:

• Neural population coding can provide a biological basis for rate-distortion theory in memory.
• The model offers a unified explanation for several aspects of visual working memory.
• This approach advances our understanding of the neural mechanisms underlying memory capacity and accuracy.