基于FPGA的基底核神经网络的实现
|
摘要
提出一种基于FPGA的基底核神经网络的实现方法。采用分段线性逼近法对原始的Izhikevich神经元数学模型进行处理,根据突触耦合原理运用DSP Builder和Simulink搭建基底核神经网络模型并进行软件仿真,运用Quarstus Ⅱ将搭建的基底核神经元网络模型下载到FPGA神经元仿真平台中并对其生物动力学放电特性进行分析。结果表明:采用分段线性逼近法完全能够实现Izhikevich神经元模型的放电特性,并且相对于原始方法节约了大量的逻辑资源;采用FPGA神经元仿真平台能够再现基底核神经网络的生物动力学特性,能够应用于大规模神经元网络的生物动力学特性研究。
This research put forward an implementation method of basal ganglia neural network based on FPGA.The mathematical model of the original Izhikevich neuron was processed by the piecewise linear approximation method.Based on the principle of synaptic coupling,a basal ganglia neural network was constructed by the DSP Builder and Simlink.The network was tested by software simulation and downloaded to the FPGA by the Quaster Ⅱ to analyze the characteristics of the biodynamics discharge.The verification results show that the discharge characteristics of Izhikevich neuron model can be fully realized by the piecewise linear approximation method,and a large amount of logic resources are saved compared with that of the original method.The biological dynamic characteristics of the basal ganglia neural network can be reproduced by the FPGA simulation platform.It is feasible to use FPGA simulation platform to implement the dynamic characteristics of a large-scale neural network.
This research put forward an implementation method of basal ganglia neural network based on FPGA.The mathematical model of the original Izhikevich neuron was processed by the piecewise linear approximation method.Based on the principle of synaptic coupling,a basal ganglia neural network was constructed by the DSP Builder and Simlink.The network was tested by software simulation and downloaded to the FPGA by the Quaster Ⅱ to analyze the characteristics of the biodynamics discharge.The verification results show that the discharge characteristics of Izhikevich neuron model can be fully realized by the piecewise linear approximation method,and a large amount of logic resources are saved compared with that of the original method.The biological dynamic characteristics of the basal ganglia neural network can be reproduced by the FPGA simulation platform.It is feasible to use FPGA simulation platform to implement the dynamic characteristics of a large-scale neural network.
引文
[1]贾军,蒋心欣,王晓民.基底神经节的结构和功能在生理学教学中的实践[J].基础医学教育,2015,17(12):1034-1035.
[2]李秀明,谢青,鲍勇,等.预测基底核性失语患者早期言语功能结局的研究[J].中国康复医学杂志,2016,31(6):659-663.
[3]SIAN J,DEXTER D T,LEES A J,et al.Alterations in glutathione levels in Parkinson's disease and other neurodegenerative disorders affecting basal ganglia[J].Annals of Neurology,1994,36(3):348-55.
[4]赵爱清.神经元网络分析与多FPGA实现[D].天津:天津职业技术师范大学,2016.
[5]SU F,WANG J,LI H Y,et al.Analysis and application of neuronal network controllability and observability[J].Chaos,2017,27(2):1063-1075.
[6]YANG S M,WANG J,LI S,et al.Cost-efficient FPGA implementation of basal ganglia and their Parkinsonian analysis[J].Neural Networks,2015,71:62-75.
[7]INDIVERI G,CHICCA E,DOUGLAS R.A VLSI array of low-power spiking neurons and bistable synapses with spiketiming dependent plasticity[C]//IEEE Transactions on Neural Networks.2006:211-221.
[8]CHEUNG K,SCHULTZ S R,LUK W.Neuro Flow:A General purpose spiking neural network simulation platform using customizable processors[J].Frontiers in Neuroscience,2016,9(19):516.
[9]YANG S,WANG J,ZHAO A,et al.A multi-FPGA embedded system for the emulation of modular small-world network with real time dynamics[C]//World Congress on Intelligent Control and Automation.2016:2198-2203.
[10]张荣华,王江.FPGA在生物神经系统模型仿真中的应用[J].计算机应用研究,2011,28(8):2949-2953.
[11]邓斌,张茂华,王晓军,等.帕金森病基底核-丘脑网络的深度脑刺激FPGA实验平台[P].中国专利:201310670414.0[P],2014-04-02.
[12]IZHIKEVICH E.Simple model of spiking neurons[J].IEEE Transactions on neural networks,2003,14(6):1569-1572.
[13]白天星.Izhikevich神经元控制参数对可塑性神经元网络动力学的影响[D].河北:河北师范大学,2015.
[14]闻佳,逯迈,董绪伟,等.基于HH模型的神经元网络的数值模拟与FPGA实现[J].航天医学与医学工程,2017,30(1):38-45.
[15]张伟.基于Izhikevich神经元模型的神经网络建模仿真[D].天津:天津医科大学,2008.
[16]孙凡,李会艳,赵爱清.多FPGA神经元网络仿真平台的设计[J].天津职业技术师范大学学报,2017,27(2):8-12.
[17]WANG J,YANG S M,DENG B,et al.Multi-FPGA implementation of feedforward network and its performance analysis[C]//Control Conference IEEE.2015:3457-3461.
[18]王美丽.基于突触可塑性的兴奋性-抑制性平衡及放电率自稳态机制研究[J].生物医学工程,2017,27(2):8-12.
[19]朱俊玲,沈强,蒋大宗.正弦电流刺激下神经元的放电模式[J].航天医学与医学工程,2002,15(2):108-111.
[2]李秀明,谢青,鲍勇,等.预测基底核性失语患者早期言语功能结局的研究[J].中国康复医学杂志,2016,31(6):659-663.
[3]SIAN J,DEXTER D T,LEES A J,et al.Alterations in glutathione levels in Parkinson's disease and other neurodegenerative disorders affecting basal ganglia[J].Annals of Neurology,1994,36(3):348-55.
[4]赵爱清.神经元网络分析与多FPGA实现[D].天津:天津职业技术师范大学,2016.
[5]SU F,WANG J,LI H Y,et al.Analysis and application of neuronal network controllability and observability[J].Chaos,2017,27(2):1063-1075.
[6]YANG S M,WANG J,LI S,et al.Cost-efficient FPGA implementation of basal ganglia and their Parkinsonian analysis[J].Neural Networks,2015,71:62-75.
[7]INDIVERI G,CHICCA E,DOUGLAS R.A VLSI array of low-power spiking neurons and bistable synapses with spiketiming dependent plasticity[C]//IEEE Transactions on Neural Networks.2006:211-221.
[8]CHEUNG K,SCHULTZ S R,LUK W.Neuro Flow:A General purpose spiking neural network simulation platform using customizable processors[J].Frontiers in Neuroscience,2016,9(19):516.
[9]YANG S,WANG J,ZHAO A,et al.A multi-FPGA embedded system for the emulation of modular small-world network with real time dynamics[C]//World Congress on Intelligent Control and Automation.2016:2198-2203.
[10]张荣华,王江.FPGA在生物神经系统模型仿真中的应用[J].计算机应用研究,2011,28(8):2949-2953.
[11]邓斌,张茂华,王晓军,等.帕金森病基底核-丘脑网络的深度脑刺激FPGA实验平台[P].中国专利:201310670414.0[P],2014-04-02.
[12]IZHIKEVICH E.Simple model of spiking neurons[J].IEEE Transactions on neural networks,2003,14(6):1569-1572.
[13]白天星.Izhikevich神经元控制参数对可塑性神经元网络动力学的影响[D].河北:河北师范大学,2015.
[14]闻佳,逯迈,董绪伟,等.基于HH模型的神经元网络的数值模拟与FPGA实现[J].航天医学与医学工程,2017,30(1):38-45.
[15]张伟.基于Izhikevich神经元模型的神经网络建模仿真[D].天津:天津医科大学,2008.
[16]孙凡,李会艳,赵爱清.多FPGA神经元网络仿真平台的设计[J].天津职业技术师范大学学报,2017,27(2):8-12.
[17]WANG J,YANG S M,DENG B,et al.Multi-FPGA implementation of feedforward network and its performance analysis[C]//Control Conference IEEE.2015:3457-3461.
[18]王美丽.基于突触可塑性的兴奋性-抑制性平衡及放电率自稳态机制研究[J].生物医学工程,2017,27(2):8-12.
[19]朱俊玲,沈强,蒋大宗.正弦电流刺激下神经元的放电模式[J].航天医学与医学工程,2002,15(2):108-111.