How Quasiconvexity works part3 Machine Learning

<p>We introduce a notion of convexity with respect to a one-dimensional operator and with this notion find a one-parameter family of different convexities that interpolates between classical convexity and quasiconvexity. We show that, for this interpolation family, the convex envelope of a continuous boundary datum in a strictly convex domain is continuous up to the boundary and is characterized as being the unique viscosity solution to the Dirichlet problem in the domain for a certain fully nonlinear partial differential equation that involves the associated operator. In addition we prove that the convex envelopes of a boundary datum constitute a one-parameter curve of functions that goes from the quasiconvex envelope to the convex envelope being continuous with respect to uniform convergence. Finally, we also show some regularity results for the convex envelopes proving that there is an analogous to a supporting hyperplane at every point and that convex envelopes are C1 if the boundary data satisfies in particular NV-condition we introduce</p> <p><a href="https://medium.com/@monocosmo77/how-quasiconvexity-works-part3-machine-learning-26bce0eef8c1">Website</a></p>