基于计算几何的非线性可视化分类器设计

A Nonlinear Visual Classifier Based on Computational Geometry

分类界面的计算是分类器设计的基本问题之一.本文以训练样本的空间表示为出发点设计了基于计算几何的区域主动生长的类界面求取方法.该方法首先对表示空间进行空间量化,并将量化后的点集按照信息表示分为基点与非基点,通过对基点区域的主动生长,使得整个表示空间任意区域均可进行类别表示,从而完成分类界面的计算过程.在具体分类器设计中,利用散点图的组合特性,将低维数据映射到多个可视空间,形成可视化组合分类器.对UCI数据集的分类实验表明,该分类器不但具有良好的可视化特性,而且分类性能已经达到或超过主流分类器的水平.

One of the basic issues in pattern recognition is to calculate the boundary between different categories.In this paper,we propose a novel method for that based on computational geometry named active expansion.At first,we quantize the description space.And then term the set as base and non-base points according their distribution,by active expanding for base points,any point in the whole space could express the category information and the boundary is obtained.Using this method,we design the scatter classifier which incorporates the active expansion with combining feature attribute of scatter plot,that mapping the data from low dimension to high dimension and conforming a visual combing classifier.The experiments against UCI datasets show that performance of the novel classifier has been equivalent to the popular classifiers,and outweigh in some dataset.