基于相对变换的半监督分类算法

为了增强基于图的局部和全部一致性(LGC)半监督算法的处理稀疏和噪声数据的能力,提出了一种基于相对变换的LGC算法。该算法通过相对变换将原始数据空间转换到相对空间,在相对空间中噪声和孤立点远离正常点,稀疏的数据变得相对密集,从而可以提高算法的性能。仿真实验结果表明,基于相对变换的LGC算法有更强的处理稀疏和噪声数据的能力。

Semi-supervised classification algorithm based on relative transformation

In order to enhance the ability of dealing with sparse and noisy data of graph-based semi-supervised learning algorithm Local and Global Consistency (LGC), a LGC algorithm based on relative transformation was proposed. The original data space was converted to the relative data space by the proposed algorithm. In the relative space, the noise and outliers would become further away from the normal points, and the sparse points would become relative closer, which can improve the performance of the LGC algorithm. The experimental results on several data sets show that the classification capability of the proposed algorithm for noisy and sparse data sets increases significantly.