Abstract: Considering that Local and Global Preserving Based Semi Supervised Dimensionality Reduction (LGSSDR) is sensitive to the selection of neighborhood parameter and inaccurate in the setting of the edge weights of neighborhood graph, a new algorithm of Local Reconstruction and Global Preserving Based Semi-Supervised Dimensionality Reduction(LRGPSSDR) was proposed in this paper. hhe algorithm can set the edge weights of neighborhood graph through minimizing the local reconstruction error and can preserve the global geometric structure of the sampled data set as well as preserving its local geometric structure. The experimental results on Extended YaleB and CML1 PIE face database demonstrate that LRGPSSDR is better than other semi-supervised dimensionality reduction algorithms in the performance of classification.

Key words: Sidcinformation, Local reconstruction, Semi-supervised learning, Dimensionality reduction