International Journal of Computer Applications (0975 – 8887) National Conference on Innovations and Recent Trends in Engineering and Technology (NCIRET-2014) 10 Role of Heterosynaptic Interaction and its effect on Development of Receptive Field Structure in Primary Visual Cortex R. P. Maheshwari Assistant Director Technical Education, Jodhpur A. R. Garg Associate Professor M.B.M. Engineering College, Jodhpur Anil Gupta Associate Professor M.B.M. Engineering College, Jodhpur ABSTRACT Many modeling studies have been performed to investigate and find a specific learning mechanism suitable or responsible for the development of simple cell receptive field structure (SCRFS). In this work, it is shown that the mechanism of spike timing dependent plasticity (STDP) when combine with heterosynaptic interaction is suitable and sufficient for development of simple cell receptive field structure. Furthermore, with this study it is confirm that in the formation of simple cell receptive field structure the required temporal and spatial relationship is provided by STDP and heterosynaptic interaction respectively. Keywords Receptive field structure, synaptic modification, Heterosynaptic interaction, 1. INTRODUCTION Experimental studies have shown that synaptic modifications induced at one synapse are accompanied by heterosynaptic changes at specific neighbouring sites that did not experience the induction activity[3,4,6,9,11]. Heterosynaptic interaction is a cellular property that has not been linked with sequence learning. Heterosynaptic competition for synapse growth or total synaptic strength has been documented at both pre and post synaptic neurons. For example, post synaptic neurons balance activity dependent potentiation of input synapse by inducing heterosynaptic depression among other input synapses, conserving the total synaptic weight on to the neurons[12]. The heterosynaptic sharing of plasticity represents a dynamic, short-term synaptic enhancement of synaptic inputs onto a common postsynaptic target. The heterosynaptic interaction changes the synapses while they were not active during the induction. Since only a fraction of the neuron’s inputs is active at a given time, or is involved in activity during a certain induction protocol, potential targets of hetero-synaptic plasticity are much more numerous. Garg, et. al[7] proposed a computational model for the formation of simple cell RF structure with the inclusion of both presynaptic and post synaptic heterosynaptic interaction. His computational model is sufficient for inputs to segregate and to maintain this segregation: starting from homogenous state to segregated ON- and OFF- inputs for the simple cell receptive field. Furthermore, there is no requirement to include additional constraints such as normalization, fixed intra-cortical synaptic strengths and hard bounds on synaptic strengths. In this work describe an integrate- and fire- neuron model has been used to each of the cortical cells. In this model spike timing dependent plasticity (STDP) is used as a learning mechanism for development of receptive field structure (RFS) in the cortical neuron in presence of inhibition. The model also incorporates heterosynaptic competition for synapse growth only for post synaptic neuron. Though, in biological system synaptic strength has been documented at both pre and post synaptic neuron. 2. MATERIAL AND METHODS The mathematical model which used for development of RFS is a set of ON-and-OFF centre Lateral Geniculate Nucleus (LGN) neurons converging on to an array of cortical neurons. The primary visual cortex (PVC) is modelled as a 2-D array of neurons. The neurons of the PVC are innervated by the ON-and OFF channels of the LGN; which are also modelled as 2-D array of neurons. For the development of thalamocortical connections a two layer structure is assumed as shown in the Figure 1. The output layer composed of a single cortical cell, which represent cell of layer IV C of cat primary visual cortex. Though in the figure there is large number of cortical neurons. The input layer, which represents the corresponding LGN layer, is subdivided into two dimensional sheets. One sheet labelled “ON” consisting of ON-type LGN cells and other sheet labeled “OFF” consisting of OFF type-LGN cells [2][10]. Cells in LGN layer are given by two dimensional position vector i, j, … etc. where, i = (i 1 ,i 2 ) and j = (j 1 ,j 2 ) and so on. Similarly the location of cortical cells are given by two dimensional position vector x,y, …. etc. where, x = (x 1 ,x 2 ) and y = (y 1 ,y 2 ) and so on. To equate the locations in the three sheets common coordinates has been used i.e. all the sheets are considered to be lying in the same spatial location.