Evaluating a Class of Dimensionality Reduction Algorithms Abstract

High-dimensional index structures (R * tree, SS-tree) are used for retrieval of multimedia objects in databases. As dimensionality increases, query performance in the index structures degrades. Dimensionality reduction algorithms are the only known solution that supports scalable object retrieval and satisfies precision of query results. By combining the methods from pattern recognition, multimedia databases, and multidimensional scaling, we are trying to evaluate a set of existing dimensionality reduction algorithms that will give us satisfying performance with respect to scalability, cost, and error rate of the output. Algorithms are applied to color and texture image vectors. Generally, there are two major research directions in dimensionality reduction. The first approach is based on distance mapping algorithms while the second is based on the SVD (singular value decomposition) technique. Then, we explore more compact ways of implementing high-dimensional indexing for large datasets.