神经网络高精度权值的模拟电路实现
ANALOG CIRCUIT IMPLEMENTATION OF NEURAL NETWORK WITH HIGH PRECISION WEIGHTS
-
摘要: 本文提出了一种电流模式4-bit可调权值模拟神经元电路,权值易存贮而且精度高。该电路可扩展为8-bit等多值分立权的神经元电路.文章扼要分析了神经元电路的工作原理,并用SPICE Ⅱ进行仿真,证明了这种电路的正确性。
-
关键词:
- 神经网络; 神经元; MOS集成电路
Abstract: A current-mode MOS neuron circuit with 4-bit programmable weights is presented using CMOS tetchnology. The weights of the neuron have high resolution and also can easily be digitally stored. The resolution can be extended into higher levels such as 8-bits, etc. by the design methodology presented in this paper. The operation principle of the neuron is discussed. Circuit simulation has been made by use of SPICE Ⅱ. The results give a good agreement for the design requirements. -
L. E. Atlas, Yoshitake Suzuki, IEEE Circuits and Devices Magazine, 5(1989)6, 20-24.[2]H. P. Graf, L. D. Jackel, IEEE Circuits and Devices Magazine, 5(1939)4, 44-49.[3]S.Y.Foo, L. R. Anderson, Yoshiyasu Takefuji, IEEE Circuits and Devices Magazine, 6(1990)4, 18-26.[4]M. Vereysen, P. G. A. Jespers, IEEE Micro, 9(198))6, 46-55.[5]A. G. Andreou et al., IEEE Trans. on NN, NN-2(1991)2, 205-213.[6]G. Moon, M. E[7]Zaghloul, R. W. Newcomb, IEEE Trans. on NN, NN-2(1992)3, 394-403.
计量
- 文章访问数: 2057
- HTML全文浏览量: 82
- PDF下载量: 560
- 被引次数: 0