电流模式神经网络分类器的CMOS实现技术

CMOS Implementation Technique of Current-mode Neural Network Classifier

  • 摘要: 运用电流模式技术,以Hamming神经网络为基本模型,进行了二值模式的最优的最小误差分类器设计实现。给出了全电流模式胜者为王(WTA)子网和模式间匹配值计算子网的CMOS实现电路,计算机仿真与理论结果相吻合。该分类器很好地模拟了生物神经网络中的侧抑制作用,所得电路结构简单,适合用标准VLSI工艺集成,有其应用价值。

     

    Abstract: The implementation of an optimum minimum error classifier based on the Hamming neural model is discussed for binary patterns in this paper by the use of current-mode technique.The CMOS implementation of current-mode winner-take all(WTA) circuit and the subnet for calculating matching scores between two patterns are demonstrated.Computer simulations are in agreement with the theory.The classifier mimics well the heavy use of lateral inhibition evident in the biological neural nets.The resulting circuits are simple and particularly suitable for integration with standard VLSI technology,thus will have perspecitve in many application areas.

     

/

返回文章
返回