Asynchronous Hardware Neural Networks with Pulse-type Chaotic Neuron Models

大脑子系统具有高度的信息处理能力，即识别和学习。但是，信息处理功能还没有被阐明。因此，为了阐明生物神经网络的信息处理功能，并将其应用于工程问题，人们对各种神经元模型和人工神经网络进行了研究。为了利用超大规模集成电路（VLSI）技术构建脑型信息处理系统，需要具有类似人脑功能的人工神经网络，本研究主要讨论了以下几个方面的内容：1.建立新的细胞体模型和活性轴突模型。2.开发新的突触模型和活性树突模型。3.构建神经网络。(1)We提出了使用脉冲型硬件神经元模型的主动轴突模型（参考文献1）。(2)We使用CMOS构造用于听觉神经网络的特征检测单元的通用型硬件神经元模型（参考文献No.3）。(3)We显影脉冲 ...更多信息 型硬件神经元器件在VDEC通过考虑电路元件分散. (1)We构建硬件有源树突模型（参考文献5）。结果表明，活性枝晶模型具有类似于生物反向传播的特性（参考文献6）。(2)We提出了使用用于塑料突触的Λ形负电阻器件的多值存储器单元的CMOS实现（参考文献No.7）。(1)We根据基底膜的生理特性，构建了一个硬件模型。结果表明，我们能够使用Neocognitron型神经网络（参考文献2）提取语音特征。(2)We提出了一个神经元爆发放电和混沌放电的最小模型，它可以被认为是Bonhoven-van der Pol（BVP）模型（参考文献4）的修改和扩展。(3)We构造了一个短时记忆电路，我们使用硬件环神经网络验证了时间模式识别电路的记忆模式（参考文献8）。(4)It阐明了具有外部输入的抑制性互耦模型可以使用具有兴奋性互耦和抑制性互耦的脉冲型硬件神经元模型来控制振荡模式。少

Brain subsystems have a high degree of information processing ability, namely recognition and learning. But, the information processing functions have not been clarified as yet. So various neuron models and artificial neural networks have been studied in order to clarify the information processing functions of biological neural networks, and apply them to engineering problems. Artificial neural networks performing similarly to the human brain are required for constructing an information processing system of brain-type using the VLSI technology.In this study, we discuss as follows :1.Develop new cell-body models and an active axon model. 2.Develop new synaptic models and an active dendrite model. 3.Construct neural networks.Results,1.(1)We propose the active axon model using pulse-type hardware neuron model (References No.1). (2)We construct a universal-type hardware neuron model using CMOS for feature-detecting cells of the auditory neural network (References No.3). (3)We develop pulse … More -type hardware neuron devices at the VDEC by considering circuit element dispersion.2.(1)We construct a hardware active dendrite model (References No.5). It is shown clearly that the active dendrite model has similar to biological backpropagation characteristics (References No.6). (2)We propose the CMOS implementation of a multiple valued memory cell using Λ-shaped negative resistance devices for plastic synapses (References No.7).3.(1)We construct a hardware model based on the physiological characteristics of basilar membrane. It is shown that we are able to extract voice features using Neocognitron-type neural networks (References No.2). (2)We propose a minimal model of neuronal bursting firing and chaotic firing that can be considered as a modification and extension of the Bonhoeffer-van der Pol(BVP) model (References No.4). (3)We construct a short-term memory circuit, and we verify the memory patterns of the temporal pattern recognition circuit using hardware ring neural networks (References No.8). (4)It is clarified that an inhibitory mutual coupling model with an external input can control oscillation modes using pulse-type hardware neuron models with excitatory mutual coupling and inhibitory mutual coupling. Less