Manifold Learning and Its Applications: Papers from the AAAI Fall Symposium
Richard Souvenir, Chair
In recent years, an impressive number of methods have been proposed for manifold learning and nonlinear dimensionality reduction. This fact illustrates both the growing interest in the area and the myriad of possible approaches to the problem. These methods vary, for example, in terms of the preservation of global or local properties of the data, regularization methods or the application of probabilistic or geometric constraints to the embedding. The resulting theory and methods of manifold learning can be applied to many areas. For example, in computer vision, most data sets are comprised of sparse, high dimensional data (for example, hundreds of images where each image contains millions of pixels). Manifold learning has been used to facilitate common computer vision tasks such as video content analysis, pose estimation, image or video segmentation, and object tracking. Similarly, applications of manifold learning are abundant in bioinformatics, natural language processing, and robotics.