Convex geometry and nonlinear approximation

A variety of properties for neural approximation follow from considerations of convexity. For example, if n and d are positive integers and X=L/sub p/([0,1]/sup d/) (with 1