Neurons can exhibit many different kinds of behaviors, such as bursting, oscillating or rebound spiking. However, research in spiking neural networks has largely focused on the neuron type known as "integrator". Recent researches have suggested that using neural networks equipped with neurons other than the integrator, might carry computational advantages. However, there still lacks an experimental validation of this idea. This study used a spiking neural network with a biologically realistic neuron model in order to provide experimental evidence on this hypothesis. The study contains two experiments. In the first experiment the optimization of the network is conducted by setting the weights to random values and then adjusting the parameters of the neurons in order to adapt the neural behaviors. In the second experiment, the parameter optimization is used in order to improve the network's performance after the weights have been trained. The results illustrate that neurons with non-standard behaviors can provide computational advantages for a network. Further implications of this study and suggestions for future research are discussed.
展开▼
