Impossibility of almost extension

From MaRDI portal




Abstract: Let (mathbfX,|cdot|mathbfX),(mathbfY,|cdot|mathbfY) be normed spaces with mathrmdim(mathbfX)=n. Bourgain's almost extension theorem asserts that for any varepsilon>0, if mathcalN is an varepsilon-net of the unit sphere of mathbfX and f:mathcalNomathbfY is 1-Lipschitz, then there exists an O(1)-Lipschitz F:mathbfXomathbfY such that |F(a)f(a)|mathbfYlesssimnvarepsilon for all ainmathcalN. We prove that this is optimal up to lower order factors, i.e., sometimes maxainmathcalN|F(a)f(a)|mathbfYgtrsimn1o(1)varepsilon for every O(1)-Lipschitz F:mathbfXomathbfY. This improves Bourgain's lower bound of maxainmathcalN|F(a)f(a)|mathbfYgtrsimncvarepsilon for some 0<c<frac12. If mathbfX=ell2n, then the approximation in the almost extension theorem can be improved to maxainmathcalN|F(a)f(a)|mathbfYlesssimsqrtnvarepsilon. We prove that this is sharp, i.e., sometimes maxainmathcalN|F(a)f(a)|mathbfYgtrsimsqrtnvarepsilon for every O(1)-Lipschitz F:ell2nomathbfY.



Cites work







This page was built for publication: Impossibility of almost extension

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2237379)