论文部分内容阅读
针对流形学习存在的对噪声敏感、易受缺失值影响问题以及现实世界数据的结构复杂性和稀疏程序大等问题,提出引入ReliefF特征估计,即应用ReliefF在流形学习中。实验分4种情况进行:一是不使用特征提取方法;二是仅使用ReliefF特征估计方法;三是仅使用有代表性的局部线性嵌入算法;四是使用改进算法。结果表明,改进算法得到的分类准确率分别比单纯使用ReliefF特征估计方法和局部线性算法都要高。
Aimed at the problems of manifold learning, such as sensitivity to noise, susceptibility to missing values, structural complexity and sparsity of real-world data, the introduction of ReliefF feature estimation is proposed, that is, ReliefF is applied to manifold learning. The experiment is divided into four cases: one is to use no feature extraction method; the other is to use ReliefF feature estimation method only; the third is to use only representative local linear embedding algorithm; the other is to use improved algorithm. The results show that the classification accuracy obtained by the improved algorithm is higher than that of the ReliefF feature estimation method and the local linear algorithm respectively.