Optimal multivalued shattering

From MaRDI portal



Abstract: We have found the most general extension of the celebrated Sauer, Perles and Shelah, Vapnik and Chervonenkis result from 0-1 sequences to k-ary codes still giving a polynomial bound. Let mathcalCsubseteq0,1,...,k−1n be a k-ary code of length n. For a subset of coordinates Ssubset1,2,...,n the projection of mathcalC to S is denoted by mathcalC|S. We say that mathcalC (i,j)-{em shatters} S if mathcalC|S contains all the 2|S| distinct vectors (codewords) with coordinates i and j. Suppose that mathcalC does not (i,j)-shatter any coordinate set of size si,jgeq1 for every 1leqi<jleqq and let p=sum(si,j−1). Using a natural induction we prove that |{mathcal C}|leq O(n^p) for any given p as noinfty and give a construction showing that this exponent is the best possible. Several open problems are mentioned.











This page was built for publication: Optimal multivalued shattering

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2910947)