A DATA MODEL FOR THE MORPHOGENETIC NEURON AND ITS ANALOGY WITH THE PHYSICAL MODEL OF BELIEF AND PLAUSIBILITY MEASURES

摘要

The classical motivation for data fusion is the presumption that high quality primary data are of paramount importance to the success of all subsequent stages. In this paper we argue against this view in this paper. This appears to be in line with recent research in neurobiology. For example the auditory place cells found in the barn owl are believed to 'compute' the azimuthal position of a sound source by detecting a time coincidence between signals which appear delayed with respect to each other by the difference of the pathlengths between the sound source and each ear. In this case two inputs control the same auditory direction cells. Two signals are fused to obtain the azimuthal dimension position. The defects of the primary signals are not really important, the importance is connected with the delay of the signals, as noted by Kosk: We note that the added value of fusion of similar sensors must originate from a non-linear combination of sensor data streams. This remark naturally gives rise to the application of novel non-linear models suggested again by neuroscience. In fact we know that the dendrite membranes in a natural neuron are not passive; the membrane conductance is voltage-dependent. The conductance is mediated by protein complexes that allow ion charges to flow across the membrane. As long ago as the 1950s Hodkin and Huxley showed how the transient changes in such conductances generate and shape the action potential. However it was at that time assumed that these conductances were limited to the axon and the adjacent cell body. We now know that many dendrites of pyramidal cells are endowed with a relative homogeneous distribution of sodium conductances as well as a diversity of calcium membrane conductances. What is the function of these active conductances? In a passive cable structure, i.e. with a voltage independent conductance, synaptic input from more distant regions of the dendritic tree would quickly saturate, delivering only a paltry amount of electrical current to the spike-initiation zone far away. If the dendritic tree is an active cable it contains sodium or calcium conductances, so the inputs in the spike-initiation zone far away, interact synergetically in a non-linear way. In this paper we present a general model of an active neuron denoted as "Morphogenetic Neuron". The morphogenetic neuron takes care of the non-linear synergetic interaction among different sensors or inputs. Finally we show that there is an analogy between the synergetic interaction given by the morphogenetic neuron potential and the physical model of the belief measure of the Dempster-Shafer theory.