Global Convergence From Random Wavelet: Benign Landscape of Sparse Blind Deconvolution in GPR
IEEE Geoscience and Remote Sensing Letters · 2026 · 23 · 1-5
DOI: 10.1109/lgrs.2026.3673217 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Geoscience and Remote Sensing Letters, 2026. Verfasst von Takanori Imai und 1 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Japan Society for the Promotion of Science Grant-in-Aid for JSPS Fellows, Council for Science, Technology and Innovation (CSTI), Cross-Ministerial Strategic Innovation Promotion Program (SIP), Third Period of SIP “Smart Infrastructur…, Japan Science and Technology Agency (JST) Fusion Oriented REsearch for Disruptive Science and Technology (FOREST) Program.
Abstract (Originalsprache)
Wavelet estimation in ground penetrating radar (GPR) through blind deconvolution has been regarded as highly initialization-dependent due to the nonconvex nature of bilinear inverse problems. This letter reveals a striking finding; sparse blind deconvolution in GPR exhibits deterministic convergence to global solutions from various wavelet initializations, eliminating the need for careful wavelet extraction. Through systematic experiments on both simulation and field-measured data, we demonstrate convergence to a single solution from diverse initializations, including Gaussian noise. We establish a quantitative connection to identifiability theory, showing that the empirical convergence boundary precisely matches the theoretical condition without adjustable parameters. Loss landscape analysis reveals that when identifiability conditions are violated, theoretically indistinguishable solutions emerge as low-loss points forming a flat valley. These findings suggest that sparse blind deconvolution in GPR is inherently well-conditioned under appropriate sparsity constraints, enabling more accessible data processing and principled hyperparameter selection.
Förderung
- Japan Society for the Promotion of Science Grant-in-Aid for JSPS Fellows
- Council for Science, Technology and Innovation (CSTI), Cross-Ministerial Strategic Innovation Promotion Program (SIP), Third Period of SIP “Smart Infrastructur…
- Japan Science and Technology Agency (JST) Fusion Oriented REsearch for Disruptive Science and Technology (FOREST) Program
Zitieren
APA
Imai, T., & Mizutani, T. (2026). Global Convergence From Random Wavelet: Benign Landscape of Sparse Blind Deconvolution in GPR. IEEE Geoscience and Remote Sensing Letters, 23, 1-5. https://doi.org/10.1109/lgrs.2026.3673217
Vancouver
Imai T, Mizutani T. Global Convergence From Random Wavelet: Benign Landscape of Sparse Blind Deconvolution in GPR. IEEE Geoscience and Remote Sensing Letters. 2026;23:1-5. doi:10.1109/lgrs.2026.3673217
BibTeX
@article{Imai2026,
author = {Imai, Takanori and Mizutani, Tsukasa},
title = {Global Convergence From Random Wavelet: Benign Landscape of Sparse Blind Deconvolution in GPR},
journal = {IEEE Geoscience and Remote Sensing Letters},
year = {2026},
volume = {23},
pages = {1--5},
doi = {10.1109/lgrs.2026.3673217}
}
EUVIMED