Data Embedding Techniques and Applications

Li Yang, Western Michigan University

As an effective way for dimensionality reduction, data embedding has direct applications in data mining, data indexing and searching, information retrieval, and multimedia data processing. As two representative techniques for data embedding, both Isomap and LLE require the construction of neighborhood graphs on which every point is connected to its neighbors. This paper reviews several techniques that have been developed to construct connected neighborhood graphs. These methods have made Isomap and LLE applicable to a wide range of data including under-sampled data and non-uniformly distributed data. Application-related issues of data embedding techniques are also discussed.