Survey and Comparison of Quality Measures for Self-Organizing Maps
Survey and Comparison of Quality Measures for Self-Organizing Maps
复制标题
自组织映射质量测量的调查和比较
DOI：
--
复制
发表时间：
2004
期刊：
影响因子：
--
通讯作者：
Georg Pölzlbauer
中科院分区：
文献类型：
--
作者：
Georg Pölzlbauer
自组织映射具有广泛的数据挖掘的有益特性，如矢量量化和投影。有几种措施可以量化这些属性中的任何一种的质量。这项工作的范围是描述和比较一些最知名的措施。这是通过进行一系列的实验，不同的地图拓扑结构与几个著名的数据集。检查这些度量是否适合确定超参数，如最佳映射大小，度量本身是否适合比较不同的映射，以及它们是否允许与类似于SOM的其他算法（例如Sammons映射）进行比较。
Self-Organizing Maps have a wide range of beneficial properties for data mining, like vector quantization and projection. Several measures exist that quantify the quality of either of these properties. The scope of this work is to describe and compare some of the most well-known measures. This is done by conducting a series of experiments for different map topologies with several well-known data sets. The measures are examined whether they are suited to determine hyperparameters like the optimal map size, how well the measure itself is suited to compare different maps, and if they allow comparison to other algorithms similar to the SOM (e.g. Sammons Mapping).