Adaptive learned iterative shrinkage-thresholding algorithm combined with a physics-driven self-supervised method for sparse spike deconvolution

Abstract

The conventional iterative shrinkage-thresholding algorithm (ISTA) faces several limitations, such as dependence on manual parameter tuning and limited ability to recover weak reflections. Hence, this study proposes a sparse spike deconvolution method based on an adaptive learned ISTA (Ada-LISTA) and a self-supervised physics-driven objective function to overcome these challenges. First, using Ada-LISTA as the backbone, the threshold and step size in the iterative process were dynamically adjusted through its adaptive parameter learning, and the seismic wavelet dictionary was used as the model input to enhance the adaptability to different complex geological scenarios. Then, a self-supervised physics-driven objective function was introduced to jointly optimize the residual of seismic records and the sparsity of reflection coefficients, further improving the interpretability of the model. Finally, comparative experiments were carried out using theoretical simulation data and actual seismic data from the Bohai Bay Basin, China, to evaluate the inversion performance of the proposed method. The experimental results indicate that, compared to the traditional ISTA algorithm, the proposed method achieved marked enhancements in reflection coefficient inversion accuracy, seismic resolution, and robustness against noise. Overall, the proposed method offers an efficient and reliable technical solution for high-resolution seismic inversion and effective recovery of weak reflection signals, providing practical support for interpreting complex subsurface geological structures.

Keywords

Sparse spike deconvolution / Seismic resolution / Weak reflection / Adaptive learned iterative shrinkage-thresholding algorithm / Physics-driven self-supervised method

Acknowledgments

None.

Funding

This work was supported in part by the Deep Earth Probe and Mineral Resources Exploration–National Science and Technology Major Project (2025ZD1004600), in part by the Natural Science Foundation of Sichuan Province (2024NSFSC0808), and in part by the Talent Project of Chengdu Technological University (2024RC028).

Conflict of interest

The authors declare they have no competing interests.

Author contributions

Conceptualization: Shunhao Hu, Shulin Pan

Formal analysis: Ziyu Qin

Investigation: Shunhao Hu, Shulin Pan, Yaojie Chen

Methodology: Yinghe Wu

Validation: Shulin Pan

Visualization: Shunhao Hu, Yaojie Chen

Writing-original draft: Shunhao Hu

Writing-review & editing: Shunhao Hu, Shulin Pan, Yinghe Wu

Availability of data

The data are available from the corresponding author on reasonable request.

References

[1] Pan S, Yan K, Lan H, Badal J, Qin Z. Adaptive step-size fast iterative shrinkage-thresholding algorithm and sparse-spike deconvolution. Comput Geosci. 2020; 134:104343. doi: 10.1016/j.cageo.2019.104343

[2] Levy S, Fullagar PK. Reconstruction of a sparse spike train from a portion of its spectrum and application to high‐resolution deconvolution. Geophysics. 1981; 46(9):1235-1243. doi: 10.1190/1.1441261

[3] Lavielle M. 2-D bayesian deconvolution. Geophysics. 1991; 56(12):2008-2018. doi: 10.1190/1.1443013

[4] Walden AT, Hosken JWJ. An investigation of the spectral properties of primary reflection coefficients. Geophys Prospect. 1985; 33(3):400-435. doi: 10.1111/j.1365-2478.1985.tb00443.x

[5] Walden AT. Non-gaussian reflectivity, entropy, and deconvolution. Geophysics. 1985; 50(12):2862. doi: 10.1190/1.1441905

[6] Symes WW. The seismic reflection inverse problem. Inverse Probl. 2009; 25(12):123008. doi: 10.1088/0266-5611/25/12/123008

[7] Chen SS, Donoho DL, Saunders MA. Atomic decomposition by basis pursuit. SIAM J Sci Comput. 1998; 20(1):33-61. doi: 10.1137/s1064827596304010

[8] Mallat SG, Zhang Z. Matching pursuits with time-frequency dictionaries. IEEE Trans Signal Process. 1993; 41(12):3397-3415. doi: 10.1109/78.258082

[9] Cai TT, Wang L. Orthogonal matching pursuit for sparse signal recovery with noise. IEEE Trans Inform Theory. 2011; 57(7):4680-4688. doi: 10.1109/tit.2011.2146090

[10] Beck A, Teboulle M. A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM J Imaging Sci. 2009; 2(1):183-202. doi: 10.1137/080716542

[11] Zhang J, Ghanem B. ISTA-Net: Interpretable Optimization- Inspired Deep Network for Image Compressive Sensing. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Salt Lake City, UT, USA: IEEE; 2018. p. 1828-1837. doi: 10.1109/cvpr.2018.00196

[12] Schnoor E, Behboodi A, Rauhut H. Generalization error bounds for iterative recovery algorithms unfolded as neural networks. Inf Inference. 2023; 12(3):2267-2299. doi: 10.1093/imaiai/iaad023

[13] Pan S, Yan K, Lan H, Badal J, Qin Z. A sparse spike deconvolution algorithm based on a recurrent neural network and the iterative shrinkage-thresholding algorithm. Energies. 2020; 13(12):3074. doi: 10.3390/en13123074

[14] Ito D, Takabe S, Wadayama T. Trainable ISTA for sparse signal recovery. IEEE Trans Signal Process. 2019; 67(12):3113-3125. doi: 10.1109/tsp.2019.2912879

[15] Hinton GE, Salakhutdinov RR. Reducing the dimensionality of data with neural networks. Science. 2006; 313(5786):504-507. doi: 10.1126/science.1127647

[16] Ranzato M, Boureau YL, LeCun Y. Sparse Feature Learning for Deep Belief Networks. In: PlattJ, KollerD, SingerY, RoweisS, editors. Proceedings of the 21st Annual Conference on Neural Information Processing Systems(NIPS 2007). Vancouver, BC, Canada: Neural Information Processing Systems; 2007. p. 1185-1192.

[17] Bengio Y, Lamblin P, Popovici D, Larochelle H. Greedy Layer-Wise Training of Deep Networks. In: Schölkopf B, Platt J, Hoffman T, editors. Proceedings of the 20th Annual Conference on Neural Information Processing Systems(NIPS 2006). Vancouver, BC, Canada: Neural Information Processing Systems Foundation; 2006. p. 153-60.

[18] Chai X, Tang G, Lin K, et al. Deep learning for multitrace sparse-spike deconvolution. Geophysics. 2021; 86(3):V207-V218. doi: 10.1190/geo2020-0342.1

[19] Gregor K, LeCun Y. Learning Fast Approximations of Sparse Coding. In: Furnkranz J, Joachims T, editors. Proceedings of the 27th International Conference on Machine Learning. Haifa, Israel; 2010. p. 399-406.

[20] Liu J, Chen X, Wang Z, Yin W. Alista:Analytic Weights are as Good as Learned Weights in LISTA. In: Proceedings of the 7th International Conference on Learning Representations (ICLR 2019). New Orleans, LA, United States: International Conference on Learning Representations (ICLR); 2019.

[21] Wu K, Guo Y, Li Z, Zhang C. Sparse Coding with Gated Learned ISTA. In: Proceedings of the 8th International Conference on Learning Representations (ICLR 2020). Addis Ababa, Ethiopia: International Conference on Learning Representations (ICLR); 2020.

[22] Zheng Z, Dai W, Xue D, Li C, Zou J, Xiong H. Hybrid ISTA: Unfolding ISTA with convergence guarantees using free-form deep neural networks. IEEE Trans Pattern Anal Mach Intell. 2022; 45(3):3226-3244. doi: 10.1109/tpami.2022.3172214

[23] Ablin P, Moreau T, Massias M, Gramfort A. Learning Step Sizes for Unfolded Sparse Coding. In: Wallach H, Larochelle H, Beygelzimer A, D’Alché-Buc F, Fox E, Garnett R, editors. Proceedings of the 33rd Annual Conference on Neural Information Processing Systems (NIPS 2019). Vancouver, BC, Canada: Neural Information Processing Systems Foundation; 2019.

[24] Li Y, Kong L, Shang F, Liu Y, Liu H, Lin Z. Learned extragradient ISTA with interpretable residual structures for sparse coding. Proc AAAI Conf Artif Intell. 2021; 35(10):8501-8509. doi: 10.1609/aaai.v35i10.17032

[25] Kim D, Park D. Element-wise adaptive thresholds for learned iterative shrinkage thresholding algorithms. IEEE Access. 2020; 8:45874-45886. doi: 10.1109/access.2020.2978237

[26] Liang W, Liu J, Zhu J. Improved analytic learned iterative shrinkage thresholding algorithm and its application to tomographic synthetic aperture radar building object height inversion. Mathematics. 2024; 12(10):1464. doi: 10.3390/math12101464

[27] Chen X, Wang Z, Liu J, Yin W. Theoretical Linear Convergence of Unfolded ISTA and its Practical Weights and Thresholds. In: Bengio S, Wallach H, Larochelle H, Grauman K, Cesa-Bianchi N, Garnett R, editors. Proceedings of the 32nd Conference on Neural Information Processing Systems. Montreal, QC, Canada: Neural Information Processing Systems Foundation; 2018. p. 9061-9071.

[28] Wang M, Zhang Z, Qiu X, Gao S, Wang Y. ATASI-Net: An efficient sparse reconstruction network for tomographic SAR imaging with adaptive threshold. IEEE Trans Geosci Remote Sens. 2023; 61:1-18. doi: 10.1109/tgrs.2023.3268132

[29] Aberdam A, Golts A, Elad M. Ada-LISTA: Learned solvers adaptive to varying models. IEEE Trans Pattern Anal Mach Intell. 2022; 44(12):9222-9235. doi: 10.1109/tpami.2021.3125041

[30] Abd Rahman AS, Elsheikh AH, Jaya MS. A deep semi-supervised learning approach for seismic reflectivity inversion. IEEE Trans Geosci Remote Sens. 2024; 62:5927714. doi: 10.1109/tgrs.2024.3401768

[31] Alfarraj M, AlRegib G. Semisupervised sequence modeling for elastic impedance inversion. Interpretation. 2019; 7(3):SE237-SE249.doi: 10.1190/int-2018-0250.1

[32] Zhao Y, Li Y, Wang S, Yang B. Physical model and super-resolution theory-guided unsupervised deep learning deconvolution for seismic resolution enhancement. IEEE Trans Geosci Remote Sens. 2025; 63:1-13. doi: 10.1109/tgrs.2025.3531359

[33] Liu B, Jiang P, Wang Q, Ren Y, Yang S, Cohn AG. Physics-driven self-supervised learning system for seismic velocity inversion. Geophysics. 2023; 88(2):R145-R161. doi: 10.1190/geo2021-0302.1

[34] Zhang J, Sun H, Zhang G, Huang X, Han L. Simultaneous physics and model-guided seismic inversion based on deep learning. IEEE Trans Geosci Remote Sens. 2024; 62:1-11. doi: 10.1109/tgrs.2024.3443970

[35] Schuster GT, Chen Y, Feng S. Review of physics-informed machine-learning inversion of geophysical data. Geophysics. 2024; 89(6):T337-T356. doi: 10.1190/geo2023-0615.1

[36] Song C, Lu M, Li Y, et al. Spatial pattern learning: Dip structure constraint multi-view convolutional neural network for pre-stacked seismic inversion. IEEE Trans Geosci Remote Sens. 2023; 61:1-16. doi: 10.1109/tgrs.2023.3307897

[37] Feng W, Liu Y, Li Y, Li H, Wang X. Acoustic impedance prediction using an attention-based dual-branch double-inversion network. Earth Sci Inform. 2025; 18(1):70. doi: 10.1007/s12145-024-01548-4

[38] Zou B, Wang Y, Chen T, Liang J, Yu G, Hu G. The domain adversarial and spatial fusion semi-supervised seismic impedance inversion. IEEE Trans Geosci Remote Sens. 2024; 62:1-15. doi: 10.1109/tgrs.2023.3336392

[39] Zhu L, Zhang F, Zhang S, Wu J. Semi-supervised intelligent inversion from prestack seismic attributes guided by geophysical prior knowledge. J Appl Geophys. 2025; 233:105620. doi: 10.1016/j.jappgeo.2025.105620

[40] Shi Y, Zhang B, Xu J, Wang Y, Jin Z. Joint supervised and semi-supervised seismic velocity model building based on VGU network. IEEE Trans Geosci Remote Sens. 2025; 63:1-14. doi: 10.1109/tgrs.2025.3533548

[41] Ge Q, Cao H, Yang Z, Yuan S, Song C. Deep-learning-based prestack seismic inversion constrained by AVO attributes. IEEE Geosci Remote Sens Lett. 2024; 21:1-5. doi: 10.1109/lgrs.2024.3373197

[42] Dai R, Yin C, Peng D. An application of elastic-net regularized linear inverse problem in seismic data inversion. Appl Sci. 2023; 13(3):1525. doi: 10.3390/app13031525

[43] Zhao X, Wu H, Li X, Peng Z, Li Y. Seismic reflection coefficient inversion using basis pursuit denoising in the joint time-frequency domain. Energies. 2020; 13(19):5025. doi: 10.3390/en13195025

[44] Chen S, Fu X, Shi Y, Cao S. Seismic wavelet shape-oriented reflectivity inversion method. J Geophys Eng. 2024; 21(4):1216-1229. doi: 10.1093/jge/gxae065

[45] Chen G, Yang W, Liu Y, Luo J, Jing H. Envelope-based sparse-constrained deconvolution for velocity model building. IEEE Trans Geosci Remote Sens. 2022; 60:1-13. doi: 10.1109/tgrs.2021.3063514

[46] Chen G, Yang W, Wang H, Zhou H, Huang X. Elastic full waveform inversion based on full-band seismic data reconstructed by dual deconvolution. IEEE Geosci Remote Sens Lett. 2022; 19:1-5. doi: 10.1109/lgrs.2022.3178915