论文部分内容阅读
流形学习算法是一种非常有效的非线性数据降维方法被广泛应用于数据挖掘,模式识别,机器学习等邻域。基于Hessian特征映射的局部线性嵌入算法(HLLE)是一种经典的流形学习算法,在很多方面都有很好的应用。但是HLLE算法对邻域的选择太过敏感,如何提高邻域选择的稳定性成了算法研究的核心内容,本文提出一种基于均值距离的选择邻域的方法,大大的提高了算法的对邻域选择的稳定性,在人工流形上取得了很好的实验效果。
Manifold learning algorithm is a very effective non-linear data dimension reduction method is widely used in data mining, pattern recognition, machine learning and other neighborhoods. The local linear embedding algorithm based on Hessian feature mapping (HLLE) is a classical manifold learning algorithm, which has good application in many aspects. However, the HLLE algorithm is too sensitive to the choice of neighborhood and how to improve the stability of the neighborhood selection becomes the core of the algorithm research. This paper presents a method of selecting neighborhoods based on mean distance, The stability of domain selection has achieved good experimental results on artificial manifold.