Isolated calmness of regularized linear inverse problems
Submitted to Optimization (2026)
Abstract
We study isolated calmness of the solution mapping of convex regularized linear inverse problems: solutions near the true signal move at most linearly with perturbations of the measurements. The main result characterizes this property through tangent cones. For convex piecewise linear-quadratic regularizers, isolated calmness turns out to be equivalent to uniqueness of the solution, which gives simple verification procedures. For analysis sparsity regularization the condition can be checked by linear programming. An example with the \(\ell_1/\ell_2\) norm shows that, beyond the polyhedral setting, a unique solution need not be isolatedly calm.
Cite
Tran T. A. Nghia, Huy N. Pham, Nghia V. Vo, and Khoa V. H. Vu. Isolated calmness of regularized linear inverse problems. Submitted to Optimization, 2026. arXiv:2609.06238.