Shi, QinfengPetterson, JamesDror, GideonLangford, JohnSmola, AlexanderVishwanathan, S.V.N.2015-12-101532-4435http://hdl.handle.net/1885/57316We propose hashing to facilitate efficient kernels. This generalizes previous work using sampling and we show a principled way to compute the kernel matrix for data streams and sparse feature spaces. Moreover, we give deviation bounds from the exact kernel matrix. This has applications to estimation on strings and graphsKeywords: Data stream; Feature space; Kernel matrices; Multi-class classification; String Kernel; Structured data; Hydraulics Graphlet kernel; Hashing; Multiclass classification; Stream; String kernel20092016-02-24