Hishiki Tetsuya, Torikai Hiroyuki
Department of Systems Innovation, OsakaUniversity, Toyonaka, Japan.
IEEE Trans Neural Netw. 2011 May;22(5):752-67. doi: 10.1109/TNN.2011.2116802.
A novel rotate-and-fire digital spiking neuron is presented. The digital neuron is a wired system of shift registers and thus it is suited to on-chip learning unlike many other analog spiking neuron models. By adjusting the wiring pattern among the registers, the digital neuron can generate spike trains with various spike patterns and can exhibit related bifurcations. A discrete-continuous hybrid map, which describes the neuron dynamics without any approximation, is derived analytically. Using the hybrid map, it is shown that the digital spiking neuron can mimic typical bifurcation phenomena and various nonlinear responses of biological neurons.
提出了一种新型的旋转发射数字脉冲神经元。该数字神经元是由移位寄存器组成的有线系统，因此与许多其他模拟脉冲神经元模型不同，它适合片上学习。通过调整寄存器之间的布线模式，数字神经元可以生成具有各种脉冲模式的脉冲序列，并能呈现相关的分岔现象。通过解析推导得到了一个离散-连续混合映射，该映射无需任何近似即可描述神经元动力学。利用该混合映射表明，数字脉冲神经元可以模拟典型的分岔现象和生物神经元的各种非线性响应。
