基于贡献率的离散Hopfield 结构优化

针对离散Hopfield 神经网络(DHNN) 结构复杂的问题, 提出一种基于贡献率的结构优化算法. 该算法利用奇异值分解方法对连接权值进行设计, 进而利用贡献率的方法对DHNN进行结构优化. 优化后的网络降低了DHNN结构的复杂程度, 使网络具有类似生物神经网络的稀疏结构, 实现了DHNN网络结构的优化. 最后, 通过水质评价和数字识别对该算法进行验证, 表明了所提出算法的有效性和可行性, 同时, 还验证了其对于大规模DHNN的有效性和适用性.

Structure optimization of discrete Hopfield based on contribution rate

To solve the problem of complex structure for the discrete Hopfield neural network(DHNN), a structural optimization algorithm based on the contribution rate is proposed. The singular value decomposition method is used to design the connection weights. On the basis of the design, the contribution rate method is adopted to prune the connection weights. The structural complexity of the DHNN is reduced after structure optimization, and it makes the DHNN with sparse network structure which is similar to biological neural network realize the structure optimization. Finally, the water quality evaluation and digital recognition are used to verify the effectiveness and feasibility of the structural optimization algorithm, and also demonstrate the effectiveness and applicability of the proposed algorithm for large scale DHNN.