新型的无师训练（General Fuzzy Min—Max，GFMM）神经网络是一种具备无师训练聚类识别能力的新型神经网络，它继承了原有GFMM网络的特点，在网络的拓扑结构和算法方面进行了较大的改进，增加了能够进行自适应在线学习的能力。基于无师训练GFMM神经网络的雷达目标识别方法完整地实现了雷达目标特征学习和识别的一体化过程。在某型对海警戒雷达舰船目标识别仿真应用实验中的结果表明：文中的方法优于其他传统的神经网络目标识别方法，在雷达目标识别方面具有良好的适用性。

In order to further improve the ability of radar target recognition of Naval vessels with neural networks, a new unsupervised general fuzzy min-max(GFMM) learning network is put forward. This network is a new neural network with the ability of unsupervised training and cluster recognition, which inherits the merits of original general min-max network and adds the self-adjusted and on-line learning capacities which improve the logical net structure and arithmetic. The method of radar target recognition based on unsupervised general min-max learning neural network perfectly accomplishes the characteristic learning of radar target in an integrated process. The results of simulated application experiment in some naval coast radar target recognition indicate that the method based on unsupervised GFMM network is better than the conventional networks and possesses the wonderful applicability in the realm of radar target recognition.

TN959.17

“十五”预先研究重点项目