Answers are grounded in this paper and cite their sources.
We study the simultaneous recovery of a spatial source term and the fractional order in a space-fractional diffusion equation from limited noisy observations. The source is represented in a finite-dimensional eigenfunction space, and the inverse problem is formulated through a two-parameter Tikhonov functional. We derive a two-time local identifiability criterion, clarify the role of the finite-dimensional parametrization, and establish existence and subsequential consistency of regularized minimizers under explicit parameter assumptions. To address the constrained optimization problem, we introduce a safeguarded sequential projected variable-step block conjugate-gradient method. The method imposes descent conditions on the actual projected displacements, recomputes the fractional-order gradient at the intermediate point ( a k + 1 , β k ) {(aᵏ⁺¹,βᵏ)} , and uses blockwise Armijo searches with negative-gradient restarts. The resulting analysis proves finite termination of both line searches, monotonic decrease of the objective functional, and block stationarity of accumulation points. Numerical experiments verify the gradient implementation, demonstrate the identifiability gain from two observation times, compare accuracy and computational cost with fixed-step baselines, and examine source-space dimension, model mismatch, regularization parameters, reference-order misspecification, and noise. The results show that the proposed method is the only tested scheme to attain the discrepancy threshold within the iteration budget while retaining a computational cost comparable to the baselines.
No takes yet. Share an insight, caveat, or question.
Xing et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: