Spherical SOM and Arrangement of Neurons Using Helix on Sphere
Spherical SOM and Arrangement of Neurons Using Helix on Sphere
复制标题
DOI：
10.2197/ipsjdc.2.133
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
Hirokazu Nishio;M. Altaf-Ul-Amin;Ken Kurokawa;S. Kanaya
中科院分区：
文献类型：
--
作者：
Hirokazu Nishio;M. Altaf-Ul-Amin;Ken Kurokawa;S. Kanaya
自组织映射（SOM）是神经网络的一种，在很多领域都有应用。在许多应用中，存在围绕神经元布置的边界。它会导致一个被称为“边界效应”的问题，因为靠近边界的神经元的邻域神经元的数量与靠近中心的神经元的数量不同。这个问题可以通过将神经元均匀地排列在球体表面上来解决。但通过传统方法，我们无法排列任意数量的神经元。因此，使用起来不方便。在这里，我们开发了一种方法来安排的神经元的任意数量的划分螺旋覆盖的表面上的一个球体分为相等的部分。与传统方法相比，它还可以更均匀地将神经元布置在球体上。
Self-organizing maps(SOM) are a kind of neural network, and are applied to many fields. In many applications, there are borders surrounding the neuron arrangement. It causes a problem which is called ‘Border effect’ because the number of neighborhood neurons of a neuron near a border is different from that of a neuron near the center. This problem can be solved by arranging neurons uniformly on the surface of a sphere. But by the conventional method we cannot arrange neurons of arbitrary number. Therefore, it is inconvenient to use. Here we developed a method to arrange the neurons of arbitrary number by dividing a helix which covers the surface of a sphere into equal parts. It also can arrange neurons on a sphere more uniformly than the conventional method.