Support Vector Density Estimation
Jason Weston, Alex Gammerman, Mark O. Stitson, Vladimir N. Vapnik, Volodya Vovk, Chris Watkins · The MIT Press eBooks · 1998
> (x); (1.1) 1. `(x) = ( 1; x ? 0 0; otherwise Generic author design sample pages 1999/07/12 15:50 2 Support Vector Density Estimation where instead of knowing the distribution function F (x) we are given the iid (independently and identically distributed) data x 1 ; : : : ; x ` (1.2) generated by F . The problem of density estimation is known to be ill-posed. "Ill-posed" means that when finding p that satisfies the equality Ap = F , where A is a linear operator, we can have large deviations in solution p corresponding to small deviations in F . In our terms, a small change in the distribution function of the continuous random variable X can cause large changes