摘要
提出一种样本点密集度的非线性流形学习算法.该算法提出了一个有效的数据点密集参数,能够很好地对非均匀数据的低维嵌入进行约束,其嵌效结果明显优于LLE算法.在人工和人脸数据集上的实验结果表明,新算法产生了较好的嵌入及分类结果.
This paper presents a non-linear manifold learning algorithm based on intensity of sample points. It proposes an effective intensity parameter of sample points, which constraints the low-dimensional embedding of uneven data well. There is a better embedding result than LLE. The experimental results on the artificial and face datasets show that the new algorithm yields a better embedding and classification result.
出处
《微电子学与计算机》
CSCD
北大核心
2012年第6期10-13,共4页
Microelectronics & Computer
关键词
流形学习
组合投资
局部线性嵌入
密集度
manifold learning
portfolio investment
Local Linear Embedding (LLE)
intensity