AN EFFICIENT COMPUTATIONAL METHOD FOR FINANCIAL AND METEROLOGICAL APPLICATIONS - Isis Project No 2486
An efficient, flexible and inexpensive method to compute Singular Value Decomposition for a large data set.
MARKETING OPPORTUNITY
The Singular Value Decomposition (SVD) or Principle Component Analysis (PCA) is the foundation of contemporary data processing for large data sets. Conventional methods for computing SVD do not extend well to environments where pluralities of computational resources are used, especially when they are running at variable speeds and have a high-bandwidth low-latency communication system. Furthermore, computing an SVD for data spanning multiple resources is expensive when using existing methods.
THE OXFORD INVENTION
Oxford inventors compute SVDs or PCAs by reducing the complex data set into several smaller computations and then combining the locally stored data sets to extract useful information. The proposed method offers several advantages such as reduction in the communication overheads, computation on resources running at different speeds and flexibility to add additional data at any time.
This work is the subject of a UK patent application, and Isis would like to talk to companies interested in commercialising this opportunity. Please contact the Isis Project Manager to discuss this further.