| Title |
Algorithms of synchronization control in neural networks / |
| Translation of Title |
Neuroninių tinklų sinchronizacijos valdymo algoritmai. |
| Authors |
Fedaravičius, Augustinas Povilas |
| DOI |
10.15388/vu.thesis.340 |
| Full Text |
|
| Pages |
192 |
| Keywords [eng] |
Neural networks ; synchronization ; control ; nonlinear dynamics |
| Abstract [eng] |
Synchronization in large populations of interacting dynamical units is the focus of intense research in physical, technological and biological systems. In neural networks, under normal conditions, it is responsible for cognition and learning, while excessive synchronization can cause abnormal brain rhythms associated with various neurological diseases. Numerous control algorithms have been developed to suppress unwanted synchronized network oscillations. It is known that an improperly designed stimulation system can damage neural tissue or the electrode itself. The mean absolute value of the stimulating current can be chosen as a performance measure for the optimization of stimulating waveform parameters in order to minimize this damage. This choice has another advantage. The optimal waveform which ensures entrainment of a spiking neuron to an external periodic stimulation is determined only by the distance between absolute extrema of the phase response curve and its amplitude. This allows to estimate the stimulation parameters empirically. The same method can be applied to a network of interacting neurons exhibiting collective periodic oscillations. The effect of high-frequency stimulation on a system of two interacting populations of QIF neurons is explained using mean-field equations averaged over the stimulation period. Such methodology can serve as an effective tool for developing stimulation algorithms to control synchronization processes in large-scale neural networks. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Doctoral thesis |
| Language |
English |
| Publication date |
2022 |
