NabhyaantraX : A Human-in-the-Loop Ensemble Learning Framework for Autonomous Exoplanet Discovery Using Space Photometry
Authors
-
Debarghya Bhowmick
Founder & Chief Executive Officer, Chief of R&D and Full Stack AI Developer (AstraGenX), Research Intern (Jadavpur University), Intern (Hindustan Aeronautics Limited, Barrackpore Division), AstraGenX, 32 H/5/2, Mahendra Roy Lane, Kolkata - 700 046, West Bengal, India
https://orcid.org/0009-0003-0468-8981
DOI:
https://doi.org/10.17010/ijcs/2026/v11/i3/176052Keywords:
Active Learning, Convolutional Neural Networks, Deep Learning, Exoplanets, Gradient Boosting, Human-in-the-Loop, Kepler, TESS, Transit Photometry, Space Informatics.Publication Chronology: Paper Submission Date : May 3, 2026 ; Paper sent back for Revision : May 9, 2026 ; Paper Acceptance Date : May 14, 2026 ; Paper Published Online : June 5, 2026.
Abstract
Since the dawn of the space-based photometric survey mission, especially legacy missions like Kepler and K2, and since the recent addition of Transiting Exoplanet Survey Satellite (TESS), an astronomical ton of time-series data has been accumulated. Although these missions have revolutionized our statistics of the planets' populations, the processes of identification of true exoplanetary transits are currently heavily limited by manual work. Ambiguities like astrophysical false positive sources (such as a grazing eclipsing binary, an instrumental false positive "signature", or blends of background sources) can only be resolved if the analysis is multi-dimensional and delicate, rather than based on a simple 1-dimensional "deterministic" piecewise response function as is often done by automated systems. These critical bottlenecks are addressed by introducing an end-to-end, human-in-the-loop (HITL) machine learning architecture that is designed to process, classify and smartly improve the accuracy of Threshold Crossing Events (TCEs) automatically. Combining an intelligently designed multi-modal ensemble of 1D Convolutional Neural Networks (CNNs) for raw morphological light curve extraction and tabular stellar metadata analysis using Gradient Boosting Decision Trees (GBDTs), NabhyaantraX generates high fidelity, state-of-the-art classification metrics. In addition, the framework incorporates a dynamic web-based visual interface, which allows domain experts to add localized astrophysical intuition, then provide real-time updates to the hyperparameters of the statistical model, and to retrain the model through an active learning graph. Detailed comparison with both the Kepler Objects of Interest (KOI) and TESS Objects of Interest (TOI) catalogues shows 22% improvement in PPFs over unimodal deep learning methods, while the number of shallow-transit Earth-like objects is recovered effectively. The basics of the system's architecture, the mathematical foundation, strict and rigorous data processing pipelines and flexible deployment capabilities, providing a scalable, interpretable and novel paradigm for the next generation of autonomous exoplanet discovery are outlined here.
Downloads
References
[1] S. Seager and G. C. Sobeck, M. H., M. Still, T. Barclay, F. Mullally, J. Troeltzsch, S. Aigrain, S. T. Bryson, D. Caldwell, W. J. Chaplin, W. D. Cochran, D. Huber, G. W. Marcy, A. Miglio, J. R. Najita, M. Smith, J. D. Twicken, and J. J. Fortney, “A unique solution of planet and star parameters from an extrasolar planet transit light curve,” Astrophysical J., vol. 585, no. 2, pp. 1038–1055, Mar. 2023, doi: 10.1086/346105.
[2] W. J. Borucki et al. "Kepler planet-detection mission: Introduction and first Results," Sci., vol. 327, no. 5968, pp. 977–980, 2010, doi: 10.1126/science.1185402.
[3] S. B. Howell, C. Sobeck, M. Haas, M. Still, T. Barclay, F. Mullally, J. Troeltzsch, S. Aigrain, S. T. Bryson, D. Caldwell, W. J. Chaplin, W. D. Cochran, D. Huber, G. W. Marcy, A. Miglio, J. R. Najita, M. Smith, J. D. Twicken, and J. J. Fortney, "The K2 Mission: Characterization and early results," Publications Astronomical Soc. Pacific, vol. 126, no. 938, 398, 2014, doi: 10.1086/676406.
[4] G. R. Ricker et al., "Transiting exoplanet survey satellite (TESS)," J. Astronomical Telescopes, Instruments, Syst., vol. 1, no. 1, 014003, 2015, doi: 10.1117/1.JATIS.1.1.014003.
[5] R. L. Akeson et al., "The NASA Exoplanet Archive: Data and tools for exoplanet research," Publications Astronomical Soc. Pacific, vol. 125, no. 989–999, Aug. 2013, doi: 10.1086/672273.
[6] J. M. Jenkins et al., "The TESS science processing operations center pipeline," Proc. SPIE, vol. 9913, id. 99133E, pp. 20, Aug. 2016, 10.1117/12.2233418.
[7] A. C. Cameron, "Astrophysical false positives in Exoplanet transit surveys," Nature, vol. 492, no. 7427, pp. 48–56, 2012, doi: 10.1038/nature11709.
[8] J. L. Coughlin et al., "Planetary candidates observed by Kepler. VII. the first fully automated catalog based on the entire 48-month data set," Astrophysical J. Suppl. Ser., vol. 224, no.1, 12, 2016, doi: 10.3847/0067-0049/224/1/12.
[9] N. M. Batalha, "Planetary candidates observed by Kepler. III. analysis of the first 16 months of data," Astrophysical J. Suppl. Ser., vol. 204, no. 2, 2013, doi: 10.1088/0067-.0049/204/2/24.
[10] A. Vanderburg and J. A. Johnson, "A Technique for Extracting Highly Precise Photometry for the Two-Wheeled Kepler Mission," Publications Astronomical Soc. Pacific, vol. 126, no. 944, 2014, doi; 10.1086/678764.
[11] C. J. Shallue and A. Vanderburg, "Identifying Exoplanets with Deep Learning: A five-planet resonant chain around kepler-80 and an eighth planet around Kepler-90," Astronomical J., vol. 155, no. 2, 94, doi: 10.3847/1538.-3881/aa9e09.
[12] M. Ansdell et al., "Scientific domain knowledge improves exoplanet transit classification with deep learning," Astrophysical J. Lett., vol. 869, no. 1, L7, 2018, doi: 10.3847/2041-8213/aaf23b.
[13] T. A. Hinners, K. Tat, and R. Thorp, "Machine Learning techniques for stellar light curve classification," Astronomical J., vol. 156, no. 1, 7, 2018, doi: 10.3847/1538-3881/aac16d.
[14] S. D. McCauliff, J. M. Jenkins, J. Catanzarite, C. J. Burke, J. L. Coughlin, J. D. Twicken, P. Tenenbaum, S. Seader, J. Li, and M. Cote et al., "Automatic classification of Kepler planetary transit candidates," Astrophysical J., vol. 806, no. 1, 6, Jun. 2015, doi: 10.1088/0004-637X/806/1/6.
[15] D. J. Armstrong, D. Pollacco, and A. Santerne," Transit shapes and self-organizing maps as a tool for ranking planetary candidates: Application to Kepler and K2,” Monthly Notices Royal Astronomical Soc., vol. 465, no. 3, pp. 2634–2642, Mar. 2017, doi: 10.1093/mnras/stw2881.
[16] H. P. Osborn, M. Ansdell, Y. Ioannou, M. Sasdelli, D. Angerhausen, D. Caldwell, J. M. Jenkins, C. Räissi, J. C. Smith, "Rapid classification of TESS planet candidates with convolutional neural networks,” Astron. Astrophys., 633, id. A53, pp. 11, Jan. 2020, doi: 10.1051/0004-6361/201935345.
[17] A. Chaushev, L. Raynard, M. R. Goad, P. Eigmüller, D. J. Armstrong, J. T. Briegal, M. R. Burleigh, S. L. Casewell, S. Gill, J. S. Jenkins, L. D. Nielsen, C. A. Watson, R. G. West, P. J. Wheatley, S. Udry, and J. I. Vines, "Classifying exoplanet candidates with Deep Learning," Monthly Notices Royal Astronomical Soc., vol. 488, no. 4, pp. 5232–5250, Oct. 2019, doi: 10.1093/mnras/stz2058.
[18] G. Olmschenk, S. I. Silva, G. Rau, R. K. Barry, E.Kruse, L. Cacciapuoti, V. Kostov, B. P. Powell, E. Wyrwas, J. D. Schnittman, and T. Barclay, "Identifying planetary transit candidates in TESS full-frame image light curves via convolutional neural networks," Astronomical J., vol. 161, no. 6, 273, May 2021, doi: 10.3847/1538-3881/abf4c6.
[19] L. Yu, A. Vanderburg, C. Huang, C. J. Shallue, I. J. M. Crossfield, B. S. Gaudi, T. Daylan, A. Dattilo, D. J. Armstrong, G. R. Ricker, R. K. Vanderspek, D. W. Latham, S. Seager, J. Dittmann, J. P. Doty, A. Glidden, and S. N. Quinn, "Identifying exoplanets with Deep Learning. III. automated triage and vetting of TESS candidates," Astronomical J., vol. 158, no. 1, 25, Jun. 2019, doi: 10.3847/1538-3881/ab21d6.
[20] B. Settles, "Active learning literature survey," Comp. Sciences Technical Rep. 1648, Univ. Wisconsin-Madison, 2009. [Online]. Available: https://minds.wisconsin.edu/handle/1793/60660
[21] J. N. Winn, "Exoplanet transits and occultations," Exoplanets, pp. 55–77. 2010. University of Arizona Press. [Online]. Available: https://arxiv.org/abs/1001.2010
[22] K. Mandel and E. Agol, "Analytic light curves for planetary transit searches," Astrophysical J. Lett., vol. 580, no. 2, L171, Nov. 2002, doi: 10.1086/345520.
[23] ASCL.net. Astrophysics Source Code Library, “Lightkurve: Kepler and TESS time series analysis in Python.” [Online]. Available: https://ascl.net/1812.013
[24] G. Kovács, S. Zucker, and T. Mazeh, "A box-fitting algorithm in the search for periodic transits," Astronomy Astrophys., vol. 391, no. 1, pp. 369–377, Jun. 2002, doi: 10.1051/0004-.6361:20020802.
[25] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE Conf. Comput. Vision Pattern Recognit., pp. 770–778, 2016, doi: 10.48550/arXiv.1512.03385.
[26] T. Chen and C. Guestrin, "XGBoost: A scalable tree boosting system," in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, pp. 785–794, Aug. 2016, doi: 10.1145/2939672.2939785.
[27] D. D. Lewis and W. A. Gale, "A sequential algorithm for training text classifiers," in Proc. 17th Annu. Int. ACM SIGIR Conf. Res. Develop. Inf. Retrieval, pp. 3–12, 1994, doi: 10.1007/978-1-4471-2099-5_1.
[28] S. Lundberg and S. I. Lee, "A unified approach to interpreting model predictions," Advances Neural Inf. Process. Syst., 3 0, 2017, doi: 10.48550/arXiv.1705.07874.
[29] S. E. Thompson et al., "Planetary candidates observed by Kepler. VIII. A fully automated catalog with measured completeness and reliability based on data release 25," Astrophysical J. Suppl. Ser., vol. 235, no. 2, 38, Apr. 2018, doi: 10.3847/1538-4365/aab4f9.
[30] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, "SMOTE: Synthetic minority over-sampling technique," J. Artif. Intell. Res., vol. 16, pp. 321–357, Jun. 2002, doi: 10.1613/jair.953.
[31] I. Loshchilov, and F. Hutter, "Decoupled weight decay regularization," Nov. 2017, doi: 10.48550/arXiv.1711.05101.
