【摘要】 文章采用神经网络动态系统中的稳定吸引子原理,设计了纠一至多位错的离散型Hopfield自反馈神经网络,其权值矩阵是主对角线为1的对称矩阵,在结构上具有较强的规律性。将这种网络用于编码,得到了SEC-DED、DEC-TED、TEC-FED非线性码。这种编码网络构造简单,所得码字种类较多,译码复杂性低。更多还原

【Abstract】 By using attractors in dynamic system of neural network, this paper designs some new discrete Hopfield neural networks with self - feedback connections, the weight matrices of which are symmetric matrices that have regular structure and take 1’ s on main diagonal. Taking advantage of these neural networks, some new sorts of codes are found. Among them are SEC -DED binary nonlinear codes, DEC - TED binary nonlinear codes and TEC - FED binary nonlinear codes with constant weight. These new neural networks are tractable in design principal. The new codes obtained from them have the advantages of lower complexity in decoding, more classes, and no checking bits.更多还原