From Steiner formulas for cones to concentration of intrinsic volumes (Q741611)

The authors develop the theory of intrinsic volumes on the unit sphere that can be interpreted as intrinsic volumes for cones. The core of the approach lies in the Steiner formula for cones and the fact that the conical intrinsic volumes sum up to one. The latter fact makes it possible to interpret the intrinsic volumes as a probability distribution on \(\{0,\dots,d\}\). The main results concern the corresponding random variable \(V_C\) for a cone \(C\), in particular, its moments, bounds for exponential moments and concentration properties. It is shown that questions about intrinsic volumes can be converted into questions about projections of a Gaussian random vector onto the cone. This approach leads to new identities and bounds for the intrinsic volumes of a cone. The results can be applied to random convex optimization problems.