| As the3rd generation of Artificial Neural Network(ANN), Spiking Neural Network has been one of the hottest research subjects of ANN recently. Initially, the learning rules of SNN were the same as the traditional neural network, which are based on mathematical method. It’s widely believed that the learning and memory in brain mainly depends on the modification mechanism in synapses among neurons. In recent years, it’s been shown that synaptic modification depends on the temporal order of the pre-and postsynaptic spiking time, which is called spike timing dependent plasticity(STDP). During the last decade, STDP has been becoming an important learning rule in SNN. Competition and Stability are two important features of STDP, which significantly affect the activity of post-synaptic neuron.For better understanding the competitive feature,2different spiking neural networks are proposed using both excitatory STDP(E-STDP) and inhibitory STDP(IN-STDP) in present study. A hybrid network with lateral inhibitory synapse connection is used to investigate the competitive behaviors of SNNs. The experiment results show that, using IN-STDP combined with dynamic learning rate, the lateral inhibitory synapse connections in the network can provide a mechanism that the neuron with the largest constant input current will win the competition, while the hybrid network can achieve this without using dynamic learning rate, using both E-STDP and IN-STDP. Meanwhile, another spiking neural network with both learning rules of E-STDP and IN-STDP is proposed to investigate the stability of STDP. The results show that IN-STDP can affect activity the post-synaptic neuron, so that the synapse between pre-and post-synaptic neurons reaches stable.In the present work, competition and stability of STDP has been investigated. Experimental results show that these two features can guide to understand properties of STDP better. Also the competition in SNN can be used in image enhancement. |
PDF Full Text Request