Gravitational Lens Parameters Estimation at Intermediate Redshifts Using Convolutional Neural Networks

Authors

  • Muhammad Doni Setiawan Astronomy Study Program, FMIPA, Institut Teknologi Bandung, Jl. Ganesa 10, Bandung 40132, Indonesia
  • Anton Timur Jaelani Astronomy Research Group and Bosscha Observatory, FMIPA, Institut Teknologi Bandung, Jl. Ganesa 10, Bandung 40132, Indonesia

DOI:

https://doi.org/10.21009/SPEKTRA.103.05

Keywords:

strong gravitational lensing, convolutional neural network, lens parameter, intermediate redshift, deep learning, SpatialDropout

Abstract

Strong gravitational lensing serves as a powerful astrophysical probe, enabling studies of dark matter, galaxy structure, and cosmological parameters. The number of strong gravitational lensing candidates at the galaxy scale is expected to reach O ~ 5 with ongoing and future wide-field galaxy surveys. Current modeling techniques largely rely on conventional fitting methods, such as least squares or maximum likelihood using Markov Chain Monte Carlo, which despite their effectiveness, are computationally expensive and require manual inspection. This motivates the development of faster yet accurate parameter estimation techniques. In this work, we construct a representative training dataset and develop an efficient Convolutional Neural Network to estimate lens parameters: the Einstein radius, axis ratio, and position angle. We utilize data from Public Data Release 3 of the Hyper Suprime-Cam Subaru Strategic Program, selecting lens galaxies in the range 0.3 ≤ z ≤ 0.9 based on the strong-lens probability distribution. We find that the choice of loss function and regularization strategy is critical. To enhance model generalization, we leverage SpatialDropout, which outperforms standard methods by addressing the spatial correlation inherent in convolutional features. Furthermore, prediction accuracy and convergence speed are strongly affected by the distribution of the training data, highlighting the importance of an appropriate loss function. Our optimized model demonstrates robust performance, achieving a Mean Absolute Error of 0.092 arcsec for the Einstein radius, providing a scalable framework for automated analysis in future wide-field surveys.

References

J. Wagner, “A model-independent characterisation of strong gravitational lensing by observables,” Universe, vol. 5, no. 177, 2019, doi: 10.3390/universe5070177.

S. Schuldt et al., “HOLISMOKES X: Comparison between neural network and semi-automated traditional modeling of strong lenses,” Astron. Astrophys., vol. 673, p. A33, 2023, doi: 10.1051/0004-6361/202244534.

C. Schaefer, M. Geiger, T. Kuntzer, and J.-P. Kneib, “Deep convolutional neural networks as strong gravitational lens detectors,” Astron. Astrophys., vol. 611, p. A2, 2018, doi: 10.1051/0004-6361/201731201.

A. Davies, S. Serjeant, and J. M. Bromley, “Using convolutional neural networks to identify gravitational lenses in astronomical images,” Mon. Not. R. Astron. Soc., vol. 487, no. 4, pp. 5263–5271, 2019, doi: 10.1093/mnras/stz1712.

R. Pearce-Casey et al., “Euclid: Searches for strong gravitational lenses using convolutional neural networks in early release observations of the Perseus field,” Astron. Astrophys., vol. 696, p. A214, 2025, doi: 10.1051/0004-6361/202453152.

Y. D. Hezaveh, L. P. Levasseur, and P. J. Marshall, “Fast automated analysis of strong gravitational lenses with convolutional neural networks,” Nature, vol. 548, pp. 555–557, 2017, doi: 10.1038/nature23463.

S. Schuldt et al., “HOLISMOKES IX: Neural network inference of strong-lens parameters and uncertainties from ground-based images,” Astron. Astrophys., vol. 671, p. A16, 2023, doi: 10.1051/0004-6361/202244325.

S. Huber and S. H. Suyu, “HOLISMOKES XII: Time-delay measurements of strongly lensed Type Ia supernovae using a long short-term memory network,” Astron. Astrophys., vol. 692, p. A15, 2024, doi: 10.1051/0004-6361/202449952.

S. Birrer and A. Amara, “lenstronomy: Multi-purpose gravitational lens modelling software package,” Phys. Dark Universe, vol. 22, pp. 189–201, 2018, doi: 10.1016/j.dark.2018.11.002.

K. C. Wong et al., “Survey of gravitationally lensed objects in HSC imaging (SuGOHI). VIII: New galaxy-scale lenses from the HSC SSP,” Publ. Astron. Soc. Jpn., vol. 74, pp. 1209–1219, 2022, doi: 10.1093/pasj/psac065.

P. Gawade et al., “Neural network prediction of model parameters for strong lensing samples from the Hyper Suprime-Cam Survey,” Mon. Not. R. Astron. Soc., vol. 540, pp. 3384–3399, 2025, doi: 10.1093/mnras/staf935.

R. C. Pinto and A. R. Tavares, “PReLU: Yet another single-layer solution to the XOR problem,” arXiv, 2024, doi: 10.48550/arXiv.2409.10821.

J. Tompson, R. Goroshin, A. Jain, Y. LeCun, and C. Bregler, “Efficient object localization using convolutional networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2015, pp. 648–656, doi: 10.1109/CVPR.2015.7298664

K. Gokcesu and H. Gokcesu, “Generalized Huber loss for robust learning and its efficient minimization for a robust statistics,” arXiv, 2021, doi: 10.48550/arXiv.2108.12627.

H. Hoshino et al., “Luminous red galaxies in clusters: Central occupation, spatial distributions and miscentring,” Mon. Not. R. Astron. Soc., vol. 452, pp. 998–1013, 2015, doi: 10.1093/mnras/stv1308.

A. T. Jaelani et al., “Survey of Gravitationally lensed Objects in HSC Imaging (SuGOHI) − X. Strong lens finding in the HSC-SSP using convolutional neural networks,” Mon. Not. R. Astron. Soc., vol. 535, pp. 1625–1639, 2024, doi: 10.1093/mnras/stae2442.

M. Tanaka, “Photometric redshift with Bayesian priors on physical properties of galaxies,” Astrophys. J., vol. 801, no. 1, p. 20, 2015, doi: 10.1088/0004-637X/801/1/20.

M. Meneghetti, Introduction to Gravitational Lensing: With Python Examples. Cham, Switzerland: Springer, 2021, doi: 10.1007/978-3-030-73582-1.

R. Kormann, P. Schneider, and M. Bartelmann, “Isothermal elliptical gravitational lens models,” Astron. Astrophys., vol. 284, pp. 285–299, 1994.

A. S. Bolton et al., “The Sloan Lens ACS Survey. V. The full ACS strong-lens sample,” Astrophys. J., vol. 682, no. 2, pp. 964–984, 2008, doi: 10.1086/589327.

C. R. Keeton, C. S. Kochanek, and U. Seljak, “Shear and ellipticity in gravitational lenses,” Astrophys. J., vol. 482, no. 2, pp. 604–620, 1997, doi: 10.1086/304172.

E. O. Ofek, H.-W. Rix, and D. Maoz, “The redshift distribution of gravitational lenses revisited: Constraints on galaxy mass evolution,” Mon. Not. R. Astron. Soc., vol. 343, no. 2, pp. 639–652, 2003, doi: 10.1046/j.1365-8711.2003.06707.x.

K. O’Shea and R. Nash, “An introduction to convolutional neural networks,” arXiv, 2015, doi: 10.48550/arXiv.1511.08458.

J. Wu, “Introduction to convolutional neural networks,” Nanjing Univ., Tech. Rep., 2017.

M. Lee, “GELU activation function in deep learning: A comprehensive mathematical analysis and performance,” arXiv, 2023, doi: 10.48550/arXiv.2305.12073.

C. Nwankpa, W. Ijomah, A. Gachagan, and S. Marshall, “Activation functions: Comparison of trends in practice and research for deep learning,” arXiv, 2018, doi: 10.48550/arXiv.1811.03378.

Downloads

Published

2025-12-30

How to Cite

Setiawan, M. D., & Jaelani, A. T. (2025). Gravitational Lens Parameters Estimation at Intermediate Redshifts Using Convolutional Neural Networks. Spektra: Jurnal Fisika Dan Aplikasinya, 10(3), 205–214. https://doi.org/10.21009/SPEKTRA.103.05

Issue

Vol. 10 No. 3 (2025): SPEKTRA: Jurnal Fisika dan Aplikasinya, Volume 10 Issue 3, December 2025

Section

Articles

License

Copyright (c) 2025 Muhammad Doni Setiawan, Anton Timur Jaelani

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

SPEKTRA: Jurnal Fisika dan Aplikasinya allow the author(s) to hold the copyright without restrictions and  allow the author(s) to retain publishing rights without restrictions. SPEKTRA: Jurnal Fisika dan Aplikasinya CC-BY or an equivalent license as the optimal license for the publication, distribution, use, and reuse of scholarly work. In developing strategy and setting priorities, SPEKTRA: Jurnal Fisika dan Aplikasinya recognize that free access is better than priced access, libre access is better than free access, and libre under CC-BY or the equivalent is better than libre under more restrictive open licenses. We should achieve what we can when we can. We should not delay achieving free in order to achieve libre, and we should not stop with free when we can achieve libre.

SPEKTRA: Jurnal Fisika dan Aplikasinya is licensed under a Creative Commons Attribution 4.0 International License.
You are free to:
Share - copy and redistribute the material in any medium or format
Adapt - remix, transform, and build upon the material for any purpose, even commercially.
The licensor cannot revoke these freedoms as long as you follow the license terms.