Tuning for criticality: a new hypothesis for sleep.
Barak A. Pearlmutter and Conor J. Houghton,
Neural Computation 21 (2009) 1622-1641.

Abstract We propose that the critical function of sleep is to prevent uncontrolled neuronal feedback while allow- ing rapid responses and prolonged retention of short- term memories. Through learning the brain is tuned to react optimally to environmental challenges. Op- timal behavior will often require rapid responses and the prolonged retention of short-term memories. At a neuronal level, these correspond to recurrent activity in local networks. Unfortunately, when a network ex- hibits recurrent activity, small changes in the param- eters or conditions can lead to runaway oscillations. Thus, the very changes that improve the processing performance of the network can put it at risk of run- away oscillation. To prevent this, stimulus-dependent network changes should only be permitted when there is a margin of safety around the current network pa- rameters. We propose that the essential role of sleep is to establish this margin by exposing the network to a variety of inputs, monitoring for erratic behav- ior, and adjusting the parameters. When sleep is not possible, an emergency mechanism must come into play, preventing runaway behavior at the expense of processing efficiency. This is tiredness.

Blurb: This paper gives a novel explanation for why we sleep: it is a proposed that sleep tunes the brain so that its network behaviour is near critical but not supercritical.