Problems where the set of discontinuities is itself an unknown in the optimization. Applications to PDEs and Optimization
Many modern optimization problems (e.g., super-resolution, optimal transport, sparse spikes) minimize an (L^2) data term plus a measure norm (total variation of a measure). This is precisely a problem on the space of Radon measures (\mathcalM(\Omega)), which is isometrically isomorphic to the dual of (C_0(\Omega)). Variational analysis in this setting uses the concept of subgradients of the total variation norm, leading to the famous "dual certificate" conditions for support recovery. Problems where the set of discontinuities is itself
The key advantage of BV over Sobolev? It allows discontinuities along lower-dimensional manifolds, making it indispensable for nonsmooth optimization. Variational analysis in this setting uses the concept
By introducing , Sobolev spaces allow us to look for solutions in a broader class of functions. This is critical for: By introducing , Sobolev spaces allow us to
In conclusion, variational analysis in Sobolev and BV spaces is a powerful tool for studying PDEs and optimization problems. The use of Sobolev and BV spaces provides a natural framework for analyzing the regularity of solutions and establishing existence and uniqueness results. The applications of variational analysis in Sobolev and BV spaces are diverse and range from image denoising to topology optimization. The MPS Siam Series on Optimization is a valuable resource for researchers and practitioners in optimization and its applications.
The celebrated ROF model for image denoising seeks (u \in BV(\Omega)) minimizing [ \frac12|u-f|_L^2^2 + \lambda |Du|(\Omega). ] The optimality condition reads (0 \in u-f + \lambda \partial |Du|(u)), i.e., the subgradient of the total variation. This leads to the nonlinear PDE [ u - f = \lambda \operatornamediv\left(\fracDuDu\right) \quad \textin the sense of distributions, ] where the right-hand side is a bounded divergence-measure field. Variational analysis guarantees existence and uniqueness, and enables primal-dual algorithms.
