Simulating of spiking neural networks

U vrijeme stalno rastuće količine prometa informacija u računalnim mrežama, potrebno je stalno pronalaziti nove načine za povećanje efikasnosti obrade podataka. To postaje posebno zahtjevno kada nam procesi nisu u potpunosti poznati ili se parametri sustava neprestano mijenjaju pa je potrebno donositi aktivne odluke o njegovoj prilagodbi. Budući da se radi o relativno slabo poznatoj grani računarstva, u radu su opisani razni modeli neuronskih mreža i njihova evolucija od početaka pa do danas. Dalje je stavljen naglasak na treću generaciju mreža – pulsne neuronske mreže. Opisani su i modeli umjetnih neurona na kojima se oni temelje. Na kraju je opisan i proces simulacije pulsnih neuronskih mreža pomoću računalne programske potpore NeuralSyns.In the days of ever growing data traffic in computer networks, there is constant need of finding new ways of increasing data efficiency of processing data. This is especially important when processes are not entirely known to us or the parameters of system are ever changing so there is need for making active decisions in regard of adapting the system. Because it is a relatively less known branch of computing science, this paper describes various neural network models and shows their evolution from the beginning up until now. In the further part emphasis is on the third generation of neural networks – spiking neural networks. Artificional neuron models which make foundation for them are also described. At the end is described process of simulation spiking neural networks with aid of software called NeuralSyns