Sparsity promoting hybrid solvers for hierarchical Bayesian inverse\n problems

The recovery of sparse generative models from few noisy measurements is an\nimportant and challenging problem. Many deterministic algorithms rely on some\nform of $\\ell_1$-$\\ell_2$ minimization to combine the computational convenience\nof the $\\ell_2$ penalty and the sparsity promotion of the $\\ell_1$. It was\nrecently shown within the Bayesian framework that sparsity promotion and\ncomputational efficiency can be attained with hierarchical models with\nconditionally Gaussian priors and gamma hyperpriors. The related Gibbs energy\nfunction is a convex functional and its minimizer, which is the MAP estimate of\nthe posterior, can be computed efficiently with the globally convergent\nIterated Alternating Sequential (IAS) algorithm \\cite{CSS}. Generalization of\nthe hyperpriors for these sparsity promoting hierarchical models to generalized\ngamma family yield either globally convex Gibbs energy functionals, or can\nexhibit local convexity for some choices for the hyperparameters. \\cite{CPrSS}.\nThe main problem in computing the MAP solution for greedy hyperpriors that\nstrongly promote sparsity is the presence of local minima. To overcome the\npremature stopping at a spurious local minimizer, we propose two hybrid\nalgorithms that first exploit the global convergence associated with gamma\nhyperpriors to arrive in a neighborhood of the unique minimizer, then adopt a\ngeneralized gamma hyperprior that promote sparsity more strongly. The\nperformance of the two algorithms is illustrated with computed examples.\n

Paper

Similar papers

© 2026 NYSGPT2525 LLC