Tomography in Fractal Neural Nets

Abstract

Tomographic procedures use multiple straight rays, running through or originating from any point P of a structure to many detector elements at the periphery. The measured specific values, are used to reconstruct the structure virtually by tracing the according rays back. In Fractal Neural Nets, whose connectivistic architecture is reflecting fractal functions, especially the squaring of complex numbers, we get the connections ordered in form of binary trees. Activation of any neuron will successively cause a wave of activities spreading over the net, approaching to the periphery. There, neurons may work as detector elements, recording the arriving sequence of activity in form of memory-strings. These memory-strings could be used any time later to activate the net in reversed direction to reconstruct the original pattern by a process, related to tomography, just not being based on straight continous rays, but on saltatory retracable sequences of neural activations.