The virtual rat: Building a computational model of the rodent whisker trigeminal system

Rodents “see” the environment mostly through their whiskers. The exquisitely sensitive, activelycontrollable whisker array receives raw sensory data in the form of mechanical signals. These signals are carried along the trigeminal pathway through a sequence of increasingly complex processing stages, from brainstem to somatosensory (“barrel”) cortex. Although many aspects of these processing stages have been characterized by a long history of experimental studies, the computational operations of the whisker-trigeminal system are poorly understood. In the present work, these computations are modelled through a goal-driven deep neural network (DNN) approach. Using a biophysically-realistic whisker array model (see companion poster, Zweifel et al., 2018), we sweep the array across a wide variety of 3D objects in highly-varying poses, angles, and speeds to generate a large dataset. Next, DNNs from …