【24h】

Adaptive-L_2 Batch Neural Gas

机译:自适应L_2批处理神经气体

获取原文

摘要

Neural Gas (NG) algorithms aim to find optimal data representations based on feature vectors. Unlike SOM, NG algorithms take into consideration the dissimilarities between prototypes in the original input space and not on a grid defined in advance. It has been successfully applied in vector quantization and clustering. However, conventional NG algorithms implicitly assume that the variables have the same importance in the clustering task. Nevertheless, some variables may be irrelevant and, among the important ones, some may be more or less important than others to the clustering task. This paper provides an adaptive batch NG algorithm that, in comparison with the traditional batch NG algorithm, has an additional step where it automatically computes the importance of the variables in the clustering task. Experiments with synthetic and real datasets show the usefulness of the proposed adaptive NG algorithm.
机译:神经气体(NG)算法旨在基于特征向量找到最佳的数据表示形式。与SOM不同,NG算法会考虑原始输入空间中而不是预先定义的网格上的原型之间的差异。它已成功应用于矢量量化和聚类。但是,常规的NG算法隐式地假设变量在聚类任务中具有相同的重要性。但是,某些变量可能无关紧要,而在重要变量中,某些变量对于聚类任​​务可能比其他变量或多或少重要。本文提供了一种自适应批处理NG算法,与传统的批处理NG算法相比,它具有一个附加步骤,可以自动计算变量在聚类任务中的重要性。通过合成数据集和真实数据集进行的实验证明了所提出的自适应NG算法的有效性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号