Application-Driven Dimension Reduction

摘要: Simplicity and eciency of linear transformations make them a popular tool for reducing dimensions (of data) before or during statistical analysis. Examples their applications include image compression reconstruction, data clustering, pattern classification, text retrieval. Linear with natural orthogonality constraints can be represented as elements Stiefel Grassmann manifolds. We advocate that the choice transformation dimension reduction is not standard; it dictated by application set, formulated an optimization problem on these above-mentioned demonstrate this idea deriving dimension-reducing in several applications, including image-based recognition objects content-based retrieval images.