Colloquium The Department of Statistics presents   Lasse Holmstrom Rolf Nevanlinna Institute
Helsinki, Finland     Reducing the Computational Complexity of Kernel Discriminant Analysis     Abstract Kernel estimate based pattern classifiers are known to discriminate well in such demanding real-world applications as optical character recognition. However, a kernel method in its basic form is often too slow to use in an on-line pattern recognition system. One can reduce the computational cost by employing radial basis function expansions that use a small number of optimized kernels. One such approach is described and comparisons with popular statstical and neural network classifiers are reported. A theoretically more tractable method is data prebinning where the data are discretized on a mesh and a kernel estimator is formed using the bin centers. We report some new results on the integrated squared error of a binned kernel density estimator and discuss its computational complexity as measured by the average number of nonempty bins.     Monday, March 30, 1998 4:10 P.M.,  1070 CEB (Duncan Hall) 4:00 P.M.: Coffee, 1044 CEB  

 
 

back to Department of Statistics Home Page This page last maintained on 03/12/98 Problems or suggestions to webmaster@stat.rice.edu Contact the Department of Statistics via stat@stat.rice.edu