Damage an‎d Recovery With Neural Networks

12/8/2018 12:00:00 AM

Due to the wide communication domain among all the network's members in complex networks, each network can be affected by disturbances which affect the entire network's performance. These disorders, depending on the type of network and its function, occurs by different reasons. Today, the study of various aspects of damage in different types of complex biological and non-biological networks has attracted a significant attention from researchers in many sciences. In this thesis, at first, the behavior and performance of the brain's biological network During the reconstruction of local damage (that occurred in brain) were examined with the help of a simplified primitive artificial neural network in frame of two different spatial-temporal dynamics. By this approach the simplified trend of two kinds of damages in brain including local-slow and local-rapid were simulated. The results and experimental data related to damages caused by brain stroke and tumors accord with each other. According to the attained results, it seems that disability is much probable to be caused by larger lesions than smaller ones. In the next section, the completed model of artificial network belonging to previous section, that is called multi-layer Presspetron neural network, was used. This time, random damages in two scheme of destruction (random vertex elimination) and (random edge elimination) were applied to small world network by which a wide spectrum of complex biological and non-biological networks can be simulated and assessed. After that, by using Presspetron neural network the caused damages were reconstructed to a significant percentage.