The rest of the approximation result is just as in the random projection case. This yields the following result whose proof is omitted as it is exactly as the proof of Theorem 4. Here the subsample is denoted as S ¯ , K ¯ is the kernel matrix defined over S ¯ and we consider...
8I]n. aIn a strsutrcutuctruerdedDHDTHTovoevrelaryla,ye,aecahcnhondoedme aminaitnaitnaisnisnifnofromrmataiotinonabaobuotutthteheOO(lo(lgong)nv) avlauleueofoof tohtehrernondoedsesand soalvnedssloolovkeus ploovkiauOp v(ilaogOn()lomgens)smageesssapgeerslpoeorkluopok. uInp.bIrnobardocaa...