采用脈沖寬度調(diào)制技術(shù)實(shí)現(xiàn)人工神經(jīng)網(wǎng)絡(luò)
The implementation of artificial neural networks using pulse width modulation technique
-
摘要: 該文提出了一種采用脈沖寬度調(diào)制(PWM)技術(shù)實(shí)現(xiàn)的人工神經(jīng)網(wǎng)絡(luò)(ANN),神經(jīng)網(wǎng)絡(luò)的權(quán)值采用數(shù)字存儲器存儲,突觸電路由開關(guān)組成,電容積分電路和電流采樣保持電路完成輸入信號和權(quán)值信號的乘加運(yùn)算,輸出采用雙電流鏡。電路結(jié)構(gòu)簡單,易于集成,可用于構(gòu)成前饋網(wǎng)絡(luò),如實(shí)現(xiàn)異或運(yùn)算的人工神經(jīng)網(wǎng)絡(luò)。Abstract: An Artificial Neural Network (ANN) is proposed which is implemented through Pulse Width Modulation (PWM) technique. The weight of the neural network is stored in digital memory, while the synapse circuits are composed of switches. The capacitor integral circuits and the current sample/hold circuits are used to implement the Weighted summation Operation (WSO) that are required in the realization of a general ANN. This circuit is easy to be integrated because of its simple structure, which could be used to construct the feed forward neural networks, for example, to implement ANN for an XOR function.
-
Yasuhiro Ota, B. M. Wilamowki, Analog implementation of pulse-coupled neural networks, IEEE Trans. on Neural Networks, 1999, 10(3), 539-544.[2]R. Coggins, M. Jabri, B. Flower, S. Pickard, A hybrid analog and digital VLSI neural network for intracardiac morphology classification, IEEE J. of Solid State Circuits, 1995, 30(5), 542-550.[3]王智,陳瓊,鄭啟倫,用BiCMOS工藝實(shí)現(xiàn)的閾值神經(jīng)元電路,華南理工大學(xué)學(xué)報(自然科學(xué)版),1999,27(12),11-15[4]陳瓊,鄭啟倫,神經(jīng)網(wǎng)絡(luò)系統(tǒng)硬件實(shí)現(xiàn)的研究進(jìn)展,電路與系統(tǒng)學(xué)報,2000,(1),48-51.[5]王若愚,鄭啟倫,黃貫光,一種基于符號邏輯的神經(jīng)元模型和電路實(shí)現(xiàn),電子科學(xué)學(xué)刊,1999,21(1),55-59[6]王伶俐,宋靈貴,吳桂初,陳偕雄,基于雙向電流型CMOS電路的譜綜合,電子科學(xué)學(xué)刊,2000,22(2),311-315.[7]張玉浩,等譯,模擬集成電路設(shè)計一電流模法,北京,高等教育出版社,1996,157-212,398-408. -
計量
- 文章訪問數(shù): 2065
- HTML全文瀏覽量: 107
- PDF下載量: 543
- 被引次數(shù): 0
下載:
