Jointly convex optimization of fold and clipping thresholds

Determine a jointly convex formulation for optimizing the contraction fold and clipping thresholds in the overload-aware product-error surrogate.

Background

The paper shows that, for fixed transform parameters, the clipping-threshold subproblem is convex, while for fixed thresholds the granular fold subproblem is a geometric program. The full joint optimization over the transform and clipping thresholds requires a nonlinear treatment, and the authors currently use alternating updates or line search rather than a jointly convex formulation.