Proposed mechanism for learning and memory erasure in a white-noise-driven sleeping cortex

Area of Science:

• Neurobiology
• Computational Neuroscience
• Information Theory

Background:

• Understanding sleep's function is a major neurobiology challenge.
• Sleep involves distinct stages, including slow-wave sleep (SWS) and rapid-eye-movement (REM) sleep.
• Neural activity patterns differ significantly across sleep stages.

Purpose of the Study:

• To model synaptic learning and memory erasure during sleep using a theory of the sleeping cortex.
• To characterize neural fluctuations during the transition from SWS to REM sleep.
• To link sleep-stage transitions to information erasure and thermodynamics.

Main Methods:

• A mean-field linearized theory of the sleeping cortex was employed.
• Statistical analysis of voltage fluctuations during SWS and REM sleep transitions.
• Application of Hebbian learning rules and information entropy calculations.

Main Results:

• Voltage fluctuations during SWS exhibit a probability density function that changes as REM sleep approaches.
• Information entropy is maximized at the SWS-to-REM sleep transition.
• Synaptic weight changes during SWS suppress neural activity, while REM sleep promotes it.

Conclusions:

• The late stage of deep SWS before REM sleep is critical for erasing labile memories.
• Information erasure during sleep transitions has an inherent entropy cost, analogous to irreversible computation.
• Sleep facilitates memory consolidation and erasure through distinct neurophysiological mechanisms.