Sparse Matrix Algorithms and Software for Information Retrieval Applications

点击翻译按钮获取中文摘要

9203004 Berry We propose the development of portable numerical software (Fortran and C) and management tools for the application of the latent semantic indexing approach to data retrieval from scientific databases. The software will implement state-of-the- art iterative numerical methods (Lanczos-based and Subspace Iteration-based) for computing the sparse singular value decomposition as needed on a full spectrum of computer systems ranging from low-end workstations (Sun-4, Machintosh II) to high- end supercomputers (CRAY Y-MP,CRAY-2S).