论文部分内容阅读
针对自组织神经网络自身的局限,将免疫克隆选择算法的克隆和变异机制引入SOM的学习算法中,提出一种免疫自组织神经网络模型,并建立了模型的学习算法。该学习算法用免疫克隆选择算法的克隆算子和变异算子改进自组织神经网络中的邻域大小和权值调整规则,使每个神经元的权值学习率和邻域大小随神经元的亲和力发生变化,从而克服了自组织神经网络分类效果受样本输入次序影响的弱点,且在很大概率上保证网络收敛到全局最优解。性能仿真结果说明该学习算法比自组织神经网络学习算法具有更好的分类准确性和泛化性能。将该模型应用雷达电子战装备的作战效能评估中,结果表明免疫自组织神经网络模型比自组织神经网络模型分类更合理。
Aiming at the limitation of self-organizing neural network, the cloning and mutation mechanism of immune clone selection algorithm is introduced into SOM learning algorithm. An immune self-organizing neural network model is proposed and a learning algorithm of the model is established. The learning algorithm uses the clonal operator and the mutation operator of the immune clonal selection algorithm to improve the neighborhood size and weight adjustment rules in the self-organizing neural network so that the weight learning rate and the neighborhood size of each neuron vary with the neuronal Thus changing the affinity so as to overcome the weakness that the self-organized neural network classification effect is affected by the input order of the sample, and to ensure that the network converges to the global optimal solution with a high probability. The performance simulation results show that the learning algorithm has better classification accuracy and generalization performance than the self-organizing neural network learning algorithm. The model is applied to the operational effectiveness evaluation of radar electronic warfare equipment. The results show that the immune self-organizing neural network model is more reasonable than the self-organizing neural network model.