An Embedding Dimension Reduction Algorithm Based on Sparse Analysis

In recent years, local manifold learning algorithms have been widely concerned, such as local linear embedding and local tangent space alignment algorithm. These algorithms are mostly based on the hypothesis of local linearization. However, the problem of whether local linearization can be realized has not been effectively solved, which makes the dimensionality reduction algorithms have poor results on natural data. In natural data, many of them are sparse, so it is important to deal with the dimension reduction for sparse data. Under the consideration of natural attributes with statistical information, an alignment of sparse local linear embedding algorithm (SLLEA) is proposed in this paper. In the algorithm, local linear range is determined dynamically according to the probability distribution of the data. For sparse data sets, the algorithm can effectively obtain local and global information. Experiments on handwork manifold and image retrieval test verify the effectiveness of the algorithm.