Sparse manifold learning and its applications in image classification
- 信息技术－会议论文 
Graph-based dimensionality reduction algorithms are important and have been commonly applied in image classification and computer vision applications. To date many approaches have been proposed, e.g. Laplacian Eigenmaps (LE), Locally Linear Embedding (LLE), Locality Preserving Projections (LPP) and ISOMAP and so on. However, all these methods need to set the k nearest neighbor parameter to address the problem. In this paper, we proposed Sparse Patch Alignment Framework to settle it. Patch Alignment Framework which unified manifold learning algorithms through two stages: local patch optimization and whole alignment. We use Sparse Coding to construct the local patch instead of using KNN, thus, the k nearest neighbor parameter is set adaptively. A lot of experiments are done to show the performance of our method. The experiment results illustrate that our method is stable and robust. Copyright 2014 ACM.