AN ANALYSIS OF CRANIOCENTRIC AND OCULOCENTRIC CODING STAGES IN A NEURAL-NETWORK MODEL OF THE SACCADIC SYSTEM

We have shown earlier in a neural network study of the saccadic system , how retinal error and an efference copy signal of eye position may g ive rise to distributed coding of target position in craniocentric coo rdinates at one level, and of motor error in oculocentric coordinates at another stage. In the present paper, the coding properties of units in the model's two hidden layers were investigated, in order to under stand at a more abstract level, how they handle their inputs and how t he two different target representations at subsequent stages emerge. I n particular, we hoped to understand better how craniocentric and ocul ocentric target representations can be constructed by merging a retino -topically coded visual signal and recruitment-coded eye position info rmation. In the first hidden layer, we found that inputs from both vis ual and oculomotor signals were nicely matched in showing similar dire ctional selectivity. Computationally, the net input of each hidden uni t can be represented by the dot product between a fixed sensitivity ve ctor, embodied by the unit's input weights, and the two-dimensional in put signal encoded by the population activity. Scaling of the resultin g total net input signal through a sigmoidal nonlinearity then yields the activity of the hidden unit. The fact that the sensitivity vectors for retinal and oculomotor signals in the first hidden layer were rou ghly aligned and matched in amplitude is the basic underlying principl e for a rough craniocentric coding at this level. Units in the second hidden layer represent motor error. This can similarly be understood o n the basis of the previously mentioned dot product characterization o f the hidden unit's connectivity. The combined process of tuned projec tion and compression by the unit's sigmoidal nonlinearity also capture s the gain-field properties of units in the first hidden layer. Our st udy suggests that the approach underlying the present analysis of an a rtificial network may be a useful tool to describe real networks and t o allow a direct comparison of simulated and neurophysiological data. Copyright (C) 1996 Elsevier Science Ltd.