基于Σ-Δ调制的比特流Sigmoid函数的实现及其在3-D空间判别网络中的应用

郭晓丹, 孟桥, 梁勇

电子学报 ›› 2015, Vol. 43 ›› Issue (5) : 862-869.

PDF(5679 KB)
PDF(5679 KB)
电子学报 ›› 2015, Vol. 43 ›› Issue (5) : 862-869. DOI: 10.3969/j.issn.0372-2112.2015.05.005
学术论文

基于Σ-Δ调制的比特流Sigmoid函数的实现及其在3-D空间判别网络中的应用

  • 郭晓丹1, 孟桥2, 梁勇2
作者信息 +

Implementation of Sigmoid Activation Function for Sigma-Delta Modulated Bit-Streams and Its Application on Network for 3-D Space Classification

  • GUO Xiao-dan1, MENG Qiao2, LIANG Yong2
Author information +
文章历史 +

摘要

本文提出一种基于Σ-Δ调制的比特流Tangent-Sigmoid(Tan-Sig)函数的数字电路实现方法.主要是通过修改反馈系数构建Σ-Δ限幅放大调制器,并以此组合实现对Tan-Sig函数的逼近.根据上述方法,在现场可编程门阵列上设计了具有非线性激活函数的比特流神经元以及复杂的比特流前馈型人工神经网络.最后利用该网络实现了在笛卡尔坐标系下的3-D空间判别.

Abstract

A digital hardware implementation of Tangent Sigmoid (Tan-Sig) activation function based on sigma-delta modulated bit-streams is proposed.Through the change of feedback coefficient in the framework of traditional sigma-delta modulator,a new limiting amplifier modulator (LAM) is fabricated,and the function approximation to Tan-Sig was achieved by the combination of several LAMs with different coefficients.Meanwhile,the bit-stream neurons with Tan-Sig activation function and the whole feed-forward artificial neural networks are implemented on field programmable gate array (FPGA).And the 3-D space classification problem in Cartesian coordinate system was solved by the neural networks presented.

关键词

比特流 / 人工神经网络 / 激活函数 / Σ-Δ调制

Key words

bit-stream / artificial neural network / activation function / sigma-delta

引用本文

导出引用
郭晓丹, 孟桥, 梁勇. 基于Σ-Δ调制的比特流Sigmoid函数的实现及其在3-D空间判别网络中的应用[J]. 电子学报, 2015, 43(5): 862-869. https://doi.org/10.3969/j.issn.0372-2112.2015.05.005
GUO Xiao-dan, MENG Qiao, LIANG Yong. Implementation of Sigmoid Activation Function for Sigma-Delta Modulated Bit-Streams and Its Application on Network for 3-D Space Classification[J]. Acta Electronica Sinica, 2015, 43(5): 862-869. https://doi.org/10.3969/j.issn.0372-2112.2015.05.005
中图分类号: TN911   

参考文献

[1] 杨国为,王守觉,卫成兵,等.基于同源的同类事物连通本性的模式分类神经网络模型[J].电子学报,2013,41(1):52-55. Yang Guo-wei,Wang Shou-jue,Wei Cheng-bing,et al.Pattern classification neural network model based on homologue connectedness[J].Acta Electronica Sinica,2013,41(1):52-55.(in Chinese)
[2] 徐宝国,宋爱国,费树岷.在线脑机接口中脑电信号的特征提取与分类方法[J].电子学报,2011,39(5):1025-1030. Xu Bao-guo,Song Ai-guo,Fei Shu-min.Feature extraction and classification of EEG in online brain-computer interface[J].Acta Electronica Sinica,2011,39(5):1025-1030.(in Chinese)
[3] Martincigh M,Abramo A.A new architecture for digital stochastic pulse-mode neurons based on the voting circuit[J].Neural Networks,IEEE Transactions on,2005,16(6):1685-1693.
[4] Chen J,Shibata T.A neuron-MOS-based VLSI implementation of pulse-coupled neural networks for image feature generation[J].Circuits and Systems I:Regular Papers,IEEE Transactions on,2010,57(6):1143-1153.
[5] Reyneri L M.A performance analysis of pulse stream neural and fuzzy computing systems[J].Circuits and Systems II:Analog and Digital Signal Processing,IEEE Transactions on,1995,42(10):642-660.
[6] Dlugosz R,Talaska T,Pedrycz W.Current-mode analog adaptive mechanism for ultra-low-power neural networks[J].Circuits and Systems II:Express Briefs,IEEE Transactions on,2011,58(1):31-35.
[7] Gatet L,Tap-Béteille H,Lescure M.Analog neural network implementation for a real-time surface classification application[J].Sensors Journal,IEEE,2008,8(8):1413-1421.
[8] Popescu S.Hardware implementation of fast neural networks using CPLD[A].Neural Network Applications in Electrical Engineering,2000.NEUREL 2000.Proceedings of the 5th Seminar on[C].Belgrade:IEEE,2000.121-124.
[9] Hariprasath S,Prabakar T N.FPGA implementation of multilayer feed forward neural network architecture using VHDL[A].Computing,Communication and Applications (ICCCA),2012 International Conference on[C].Dindigul,Tamilnadu:IEEE,2012.1-6.
[10] Cheung K F,Tang P Y H.Sigma-delta modulation neural networks[A].Neural Networks,1993.,IEEE International Conference on[C].San Francisco,CA:IEEE,1993.489-493.
[11] Murahashi Y,Hotta H,Doki S,et al.Pulsed neural networks based on delta-sigma modulation suitable for hardware implementation[A].Neural Networks,2004.Proceedings.2004 IEEE International Joint Conference on[C].Budabest,Hungary.IEEE,2004.4:2607-2612.
[12] Hovin M,Wisland D,Berg Y,et al.Delta-Sigma modulation in single neurons[A].Circuits and Systems,2002.ISCAS 2002.IEEE International Symposium on[C].Scottsdale,AZ:IEEE,2002.5:V-617-V-620.
[13] Liang Yong,Wang Zhi-gong,Meng Qiao,et al.Bit-stream linear artificial neural networks based on Sigma-delta modulation[J].High Technology Letters:English Edition,2012,18(2):120-123.
[14] Liang Y,Wang Z,Meng Q,et al.FPGA implementation of bit-stream neuron and perceptron based on sigma delta modulation[J].Journal of Southeast University,2012,28(3):282-286.
[15] Machul O,Hammerschmidt D,Weiler D,et al.Nonlinear function generation using oversampled Σ-Δ-modulators[A].Circuits and Systems,2000.Proceedings.ISCAS 2000 Geneva.The 2000 IEEE International Symposium on[C].Geneva:IEEE,2000,1.611-614.
[16] Hidaka Y,Fujisaka H,Sakamoto M,et al.Piecewise linear operations on sigma-delta modulated signals[A].Electronics,Circuits and Systems,2002.9th International Conference on[C].Dubrovnik:IEEE,2002.3,983-986.
[17] Hayashi K,Katao T,Fujisaka H,et al.Piecewise linear circuits operating on first-order multi-level and second-order binary sigma-delta modulated signals[A].Circuit Theory and Design,2007.ECCTD 2007.18th European Conference on[C].Seville:IEEE,2007.683-686.
[18] Fujisaka H,Kurata R,Sakamoto M,et al.Bit-stream signal processing and its application to communication systems[J].IEE Proceedings-Circuits,Devices and Systems,2002,149(3):159-166.
[19] Hikawa H.Frequency-based multilayer neural network with on-chip learning and enhanced neuron characteristics[J].Neural Networks,IEEE Transactions on,1999,10(3):545-553.

基金

国家自然科学基金 (No.60576028)

PDF(5679 KB)

1961

Accesses

0

Citation

Detail

段落导航
相关文章

/