摘要： 为解决目前数字型演化硬件研究中存在的电路编码困难问题，提出一个可用矩阵形式描述组合电路的类神经网络门级电路模型，讨论在此模型上进行电路编码的具体方法。根据编码矩阵特点，对标准遗传算法进行改进，设计遗传操作算子、适应度评估方法等。通过无刷直流电动机电子换相电路的成功演化实例，验证了采用矩阵编码和改进遗传算法实现数字电路演化的可行性。

关键词: 神经网络, 组合电路, 演化硬件

Abstract: For the purpose of solving the encoding problem harassed the digital Evolvable Hardware(EHW) researchers, a gate-level circuit model which is based on the similarities between combinatorial circuit and neural network is proposed, on which the matrix encoding scheme of combinatorial circuit is discussed. An improved genetic algorithm is used to evolve the encoding matrix, genetic operators and fitness evaluation method are designed according to the characteristics of circuit encoding. The implementation of the commutation circuit of brushless direct current motor proves the feasibility of the implementation method of digital EHW by the using of matrix encoding scheme and the improved genetic algorithm.

Key words: neural network, combinatorial circuit, Evolvable Hardware(EHW)

中图分类号: 

• TP301.6