Applied Mathematics & Information Sciences

Author Country (or Countries)

P. R. China


The deficiency of the ability for preserving global geometric structure information of data is the main problem of existing semi-supervised dimensionality reduction with pairwise constraints. A dimensionality reduction algorithm called Semi-supervised Sparsity Pairwise Constraint Preserving Projections based on Genetic Algorithm (SSPCPPGA) is proposed. On the one hand, the algorithm fuses unsupervised sparse reconstruction feature information and supervised pairwise constraint feature information in the process of dimensionality reduction, preserving geometric structure in samples and constraint relation of samples simultaneously. On the other hand , the algorithm introduces the genetic algorithm to set automatically the weighted trade-off parameter for full fusion. Experiments operated on real world datasets show, in contrast to the existing typical semi-supervised dimensionality reduction algorithms with pairwise constraints and other semi-supervised dimensionality reduction algorithms on sparse representation, the proposed algorithm is more efficient.

Suggested Reviewers