优先排序神经网络K网覆盖分类研究

Research on K-classification covering for PONN

从高维空间特征点覆盖的角度,讨论了优先度排序神经网络(PONN)算法,提出了非各向同性的K网覆盖算法(KPA)算法,最后给出标准测试集和应用测试集的比较结果,并对其与各向同性覆盖中心适配选择算法(CASA)进行了分析与比较,实验结果表明KPA算法在样本连续性构造方面优于CASA算法。

From the aspect of coverage in the High-Dimensional Space （HDS）, we discussed algorithms of Priority Ordered Neural Network （PONN）, and promoted K-Partitioning Algorithm （KPA） based on anisotropy in HDS. Benchmark testing and application were made on KPA as well as the comparison with Center Adaptive Selection Algorithm （ CASA）. The resuhs of experiment prove that the constructive method of KPA is superior to CASA especially in continuous samples.