数字逻辑的稳健神经网络实现

多层二进前向神经网络或布尔神经网络作为典型的人工神经网络模型，研究和应用的十分广泛，这里在分析数字逻辑基本运算和神经元关系后，提出了一种改进的利用三层前向感知器神经网络实现任意数字逻辑函数的新算法，该算法由稳健的感知器构造神经网络，并引入汉明距离化简、卡诺图化简和最小项抑制来降低网络的复杂性，由此算法构造的神经网络不但具有稳健性能，而且消除了对数字输入变量所作的变换，使其更加简单、规范，容错能力更强，可广泛应用于对数字电路设计、编码密码的研究。

Realization of robust neural network for digital logic

Binary feedforward neural network or Boolean neural network as a typical neural network model was widely studied and applied. After analyzing the relationship between basic operations of digital logic and a neuron, an improved algorithm for implementing digital logic based on three-layer feedforward perceptron neural network is proposed in the paper. The network utilized here is based on the robust perceptrons and its complexity is reduced by Hamming distance simplification, Kaunaugh map simplification and minterm suppression. So the neural network structured has robustness capability and the transform of input digital variables is removed. The network becomes simpler and has stronger fault-tolerated capability. The result can be widely used in the study of digital circuits design, encode and cipher.