This award supports Drs. Patrice Simard, Vladimir Vapnik, Yann LeCun, and Leon Bottou to collaborate with Klaus-Robert M¼ller of GMD FIRST (Research Center for Information Technology) in Berlin. The collaboration will extend current knowledge about support vector machines (SVMs) which can represent functions in spaces with trillions of dimensions and are used for such applications as time-series prediction, image analysis, and optical character recognition. Specifically, the research will study methods for incorporating invariance properties into SVMs, methods for automatically choosing the parameters and kernels of SVMs, numerical techniques for computing support vectors, extending SVM beyond current uses, and improving pattern-recognition performance. One result of the research will be a numerical code for the computation of support vectors that will be available to the research community.
