Neural network model of the cerebellum: temporal discrimination and the timing of motor responses

• Authors: Dean V. Buonomano;Michael D. Mauk
• Venue: Neural Computation
• Year: 1994

Substantial evidence has established that the cerebellum playsan important role in the generation of movements. An importantaspect of motor output is its timing in relation to externalstimuli or to other components of a movement. Previous studiessuggest that the cerebellum plays a role in the timing ofmovements. Here we describe a neural network model based on thesynaptic organization of the cerebellum that can generate timedresponses in the range of tens of milliseconds to seconds. Incontrast to previous models, temporal coding emerges from thedynamics of the cerebellar circuitry and depends neither onconduction delays, arrays of elements with different timeconstants, nor populations of elements oscillating at differentfrequencies. Instead, time is extracted from the instantaneousgranule cell population vector. The subset of active granule cellsis time-varying due to the granule---Golgi---granule cell negativefeedback. We demonstrate that the population vector of simulatedgranule cell activity exhibits dynamic, nonperiodic trajectories inresponse to a periodic input. With time encoded in this manner, theoutput of the network at a particular interval following the onsetof a stimulus can be altered selectively by changing the strengthof granule ’ Purkinje cell connections for those granulecells that are active during the target time window. The memory ofthe reinforcement at that interval is subsequently expressed as achange in Purkinje cell activity that is appropriately timed withrespect to stimulus onset. Thus, the present model demonstratesthat a network based on cerebellar circuitry can learnappropriately timed responses by encoding time as the populationvector of granule cell activity.