Real-world-time simulation of memory consolidation in a large-scale cerebellar model

• Main Authors: Masato eGosui, Tadashi eYamazaki
• Format: Article
• Language: English
• Published: Frontiers Media S.A. 2016-03-01
• Series: Frontiers in Neuroanatomy
• Subjects: Cerebellum; memory consolidation; Model; Realtime Simulation; Optokinetic Response; Graphics processing unit
• Online Access: http://journal.frontiersin.org/Journal/10.3389/fnana.2016.00021/full

We report development of a large-scale spiking network model of thecerebellum composed of more than 1 million neurons. The model isimplemented on graphics processing units (GPUs), which are dedicatedhardware for parallel computing. Using 4 GPUs simultaneously, we achieve realtime simulation, in which computer simulation ofcerebellar activity for 1 sec completes within 1 sec in thereal-world time, with temporal resolution of 1 msec.This allows us to carry out a very long-term computer simulationof cerebellar activity in a practical time with millisecond temporalresolution. Using the model, we carry out computer simulationof long-term gain adaptation of optokinetic response (OKR) eye movementsfor 5 days aimed to study the neural mechanisms of posttraining memoryconsolidation. The simulation results are consistent with animal experimentsand our theory of posttraining memory consolidation. These resultssuggest that realtime computing provides a useful means to studya very slow neural process such as memory consolidation in the brain.