【摘要】 针对Hopfield神经网络的自联想特性,提出一种新的带有粒子群优化过程的Hopfield分类算法(PSO-HOP)。该算法采用了Blatt-Vergin(BV)学习算法,一定程度上克服了传统Hopfield容量低的特点。与此同时,还提出了先测量后训练的方法来降低算法的复杂度,提高分类效率,并探讨了样本属性以及类标号在Hopfield神经网络的表示方法,使其能够很好地处理空缺值等噪声数据。通过采用离散型粒子群优化算法对Hopfield的拓扑结构进行优化,可以将多余的神经元分配给不同属性,使得属性在分类中的权重发生改变,从而提高分类精度,避免陷入局部最优值。从统计不同属性被分配神经元的次数中,可以反映出不同属性的重要程度。从大量实验结果可以看出,该算法具有较高的鲁棒性和分类准确度。更多还原

【Abstract】 According to Hopfield neural network’s auto-associate property,a new Hopfield classification algorithm(PSO-HOP for short) was proposed,which includes the discrete particle swarm optimization.This new classification algorithm includes the learning algorithm,which was proposed by Blatt and Vergin,to train the neural network,enabling the neural network to overcome the demerit of low pattern capability.In addition,a measure-train method was put forward to decrease the time complexity,making the classification more efficient.To better deal with the noise data such as missing value,the algorithm takes the advantage of the auto-associate property of the discrete Hopfield neural network,and applies it in the expression of the attribute and class.The PSO-HOP algorithm also contains a discrete particle swarm optimization to optimize the structure of the neural network.By distributing the redundant neurons to the attributes,the weight of the attributes can be changed,resulting in the increase of the precision and reflecting the significance of different attributes.The PSO-HOP algorithm is a new approach to classify the data and its validity is substantiated by abundant experiment results更多还原