Dimensionality Reduction With Extreme Learning Machine Based On Manifold Preserving

As a nonlinear dimensionality reduction technique, manifold learning method is used to solve the computational complexity of high-dimensional data, which has been widely used in the visualization and feature extraction. However, most of existing methods are sensitive to noise and usually cannot get the accurate projection function, which limit its practical application. To overcome these problems, we propose a nonlinear dimensionality reduction method with Extreme Learning Machine based on Manifold Preserving called MP-ELM. MP-ELM takes both manifold structure and distance information into account, which can heighten dimensionality reduction effect and enhance noise immunity. The results of the visualization experiment on artificial data, clustering and face recognition experiment on several real-word database show that our method significantly outperforms the compared dimensionality reduction methods.