Learning in the Vestibular System: Simulations of Vestibular Compensation Using Recurrent Back-Propagation

Part of Advances in Neural Information Processing Systems 4 (NIPS 1991)

Bibtex Metadata Paper


Thomas Anastasio


Vestibular compensation is the process whereby normal functioning is regained following destruction of one member of the pair of peripheral vestibular receptors. Compensation was simulated by lesioning a dynamic neural network model of the vestibulo~ular reflex (VOR) and retraining it using recurrent back-propagation. The model reproduced the pattern of VOR neuron activity experimentally observed in compensated animals, but only if connections heretofore considered uninvolved were allowed to be plastic. Because the model incorporated nonlinear units, it was able to reconcile previously conflicting, linear analyses of experimental results on the dynamic properties of VOR neurons in normal and compensated animals.


Vestibular compensation is one of the oldest and most well studied paradigms in motor learning. Although it is neurophysiologically well described, the adaptive mechanisms underlying vestibular compensation, and its effects on the dynamics of vestibular responses, are still poorly understood. The purpose of this study is to gain insight into the compensatory process by simulating it as learning in a recurrent neural network model of the vestibulo-ocular reflex (VOR).