pith. sign in

arxiv: 2605.27527 · v2 · pith:K4GCSEALnew · submitted 2026-05-26 · 🌌 astro-ph.IM · cs.LG

Probabilistic Data-Driven Modelling of Astrophysical Transients: The Neural Process Family for Ultrafast and Class-Agnostic Light Curve Reconstruction with NightLANP

Siddharth Chaini , Federica B. Bianco , Ashish Mahabal This is my paper

Pith reviewed 2026-07-01 16:01 UTC · model grok-4.3

classification 🌌 astro-ph.IM cs.LG
keywords neural processeslight curve reconstructionastrophysical transientsRubin ObservatoryGaussian processesmeta-learningprobabilistic interpolationamortized inference
0
0 comments X

The pith

Attentive Neural Processes outperform Gaussian Processes and neural networks for sparse multi-band light curve reconstruction while running orders of magnitude faster.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that meta-learning an Attentive Neural Process on simulated transients yields a single class-agnostic model for rapid amortized inference on light curve interpolation. This combines the probabilistic framework of Gaussian Processes with the scalability of deep learning by shifting the bulk of computation to training time. A sympathetic reader would care because the Vera C. Rubin Observatory will produce vast amounts of sparse irregular data across six bands where per-curve fitting methods scale poorly. The model handles all bands simultaneously and produces well-calibrated uncertainties that avoid the overconfidence typical of standard neural networks and the underconfidence of Gaussian Processes.

Core claim

By meta-learning on diverse simulated transients, Attentive Neural Processes enable ultrafast class-agnostic interpolation of light curves that outperforms a suite of Gaussian Processes and neural networks on every tested metric spanning regression quality, astrophysical feature recovery, and probabilistic calibration, while completing all-band interpolation in microseconds.

What carries the argument

The Attentive Neural Process, which performs meta-learning to enable amortized probabilistic inference across multiple transient classes without per-curve optimization or kernel specification.

If this is right

  • Simultaneous interpolation of all six bands in microseconds suitable for real-time processing of the Rubin alert stream.
  • Superior performance on every tested metric including regression, feature recovery, and uncertainty calibration compared to benchmarks.
  • Delivery of sharp well-calibrated uncertainties that avoid overconfidence of neural networks and underconfidence of Gaussian Processes.
  • Scalable foundation for transient science without the need for individual light curve fitting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Application to other domains with sparse irregular time-series data could benefit from similar amortized probabilistic models.
  • Integration with downstream tasks like classification or parameter estimation might further improve efficiency in large surveys.
  • Extension to include additional observational constraints or real-time updating could enhance adaptability to varying conditions.

Load-bearing premise

Meta-learning on diverse simulated transients produces a class-agnostic model that generalizes to real Rubin observations without requiring domain-specific kernel choices or per-curve fitting.

What would settle it

A direct comparison on actual observed Rubin light curves where the Attentive Neural Process fails to outperform the benchmark methods on regression or calibration metrics would falsify the performance claims.

Figures

Figures reproduced from arXiv: 2605.27527 by Ashish Mahabal, Federica B. Bianco, Siddharth Chaini.

Figure 1
Figure 1. Figure 1: Schematic overview of the meta-learning training procedure for our Attentive Neural Process (ANP) model on light curves. (a) We start with a ground truth set of 15,000 (2- or 10-day cadence) simulated light curves spanning 15 different transient classes (1,000 light curves per class). (b) Meta-training involves choosing a batch of light curves at every iteration and randomly drawing between 12 and 60 point… view at source ↗
Figure 2
Figure 2. Figure 2: Model architecture of the Attentive Neural Pro￾cess, adapted from Kim et al. (2019). The encoder consists of two parallel pathways, a deterministic and a latent path, described in detail in Section. 2. The decoder concatenates the target location (⃗t∗, ⃗b∗), the deterministic representation r⃗∗, and the latent sample z⃗, and passes them through an MLP to produce the predictive distribution for F⃗ ∗. The cr… view at source ↗
Figure 3
Figure 3. Figure 3: An example of a Type IIP Supernova from our synthetic dataset in the 6 LSST bands (ugrizy). The full set of simulated points are denoted by crosses (target points), while a realistic subset that would be observed by LSST is de￾noted by the circles (context points). Note that this is just a single realization of an LSST cadence based on LSST OpSims v5.1.1. The task of light curve interpolation is to reconst… view at source ↗
Figure 4
Figure 4. Figure 4: Model predictions for the example Type IIP supernova in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: As [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Summary of the performance of the ANP and seven benchmark models for all metrics considered in this work (see Section. 3). The inset percentage (win rate) specifies the percentage of light curves for which the specified model is the best performer as per that metric (higher is better). ANP is overwhelmingly the best model for all reconstruction metrics (top 9 rows) and when considering the prediction of ti… view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of per light curve inference times (in seconds, log y-axis) for ANP and all seven benchmark models considered in this work (Section. 3), excluding training time (lower is better). ANP achieves ∼10−6 s/light curve – four orders of magnitude faster than the next-best neural benchmark (MLP) and over five orders of magnitude faster than GPs [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Results I: Distribution of four regression quality metrics (Section. 3.2.1) for the unseen 3,750 light curves (15 transient classes, 250 objects each) across all eight reconstruction models (lower is better). Each light curve includes context points for 10 realizations based on the expected LSST cadence. (a) Mean Squared Error (MSE) is computed per object across all bands, while (b) Mean Squared Scaled Err… view at source ↗
Figure 9
Figure 9. Figure 9: Results II: As [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Results III: Distribution of four peak estimation metrics (see Section. 3.2.3), evaluating how accurately each model recovers the time and magnitude of peak brightness (lower is better). Like in [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Results IV: Distribution of peak-time errors like [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Results V: Distribution of two probabilistic eval￾uation metrics (see Section. 3.2.4). Like in [PITH_FULL_IMAGE:figures/full_fig_p016_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Results VI: Confidence calibration plot for ANP and all seven benchmark models considered in this work (see Section. 3). The top panel displays the empirical coverage (PICP, see Section. 3.2.4) as a function of the predicted confi￾dence interval, where the solid black diagonal line represents perfect statistical calibration. The bottom panel shows the calibration residuals (empirical PICP minus predicted … view at source ↗
Figure 14
Figure 14. Figure 14: Results VII: Distribution of two probabilistic evaluation metrics, based on the Gneiting et al. (2007) frame￾work (lower is better). Like in [PITH_FULL_IMAGE:figures/full_fig_p018_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Probability Integral Transform histograms (PIT, Gneiting et al. 2007) for all eight models considered in this work. For a perfectly calibrated model, PIT values follow a Uniform(0, 1) distribution, producing a flat histogram (red dashed line at density = 1). PIT values are calculated for all points across all light curves and across all realizations. The KS statistic (lower is better; derived from the Kol… view at source ↗
Figure 16
Figure 16. Figure 16: Results VIII: Distribution of the Mean Squared Scaled Error as a function of the physical class; SN-like transients are modelled most effectively by ANP. For thermonuclear SNe, which are a more homogeneous set of SNe, ANP still outperforms other models but with a less dramatic difference. For Other transients, which include ILOT and CART (see Section. 2.3), the ANP predictions are statistically as good as… view at source ↗
read the original abstract

Astrophysical observations from Earth are subject to weather, environmental, and scientific constraints that lead to sparse, irregular light curves. On the eve of the Vera C. Rubin Observatory Legacy Survey of Space and Time, its dataset offers unprecedented opportunities for transient science. Yet a key challenge remains its cadence, sparse and irregular across six bands, limiting inference. Interpolation helps mitigate this, with Gaussian Processes the standard, but they struggle with cross-band correlations, require a priori kernel specification, and must be fit to each light curve individually, hence scaling poorly. Here, we introduce the neural process family for light curve reconstruction, combining the probabilistic framework of Gaussian Processes with the scalability of deep learning. By meta-learning on diverse simulated transients, Attentive Neural Processes shift the bulk of computation to training, enabling rapid, amortized inference with a class-agnostic model. Evaluated on realistic Rubin cadences across 15 transient classes, we show that even an unoptimized, out-of-the-box Attentive Neural Process consistently outperforms all benchmarks -- a suite of Gaussian Processes and neural networks -- on every tested metric, spanning regression quality, astrophysical feature recovery, and probabilistic calibration. Our model interpolates all bands simultaneously in microseconds, over four orders of magnitude faster than the next-best neural benchmark and five faster than Gaussian Processes, demonstrating the potential of neural processes for the nightly Rubin alert stream. Attentive Neural Processes avoid the overconfidence of standard neural networks and the underconfidence of Gaussian Processes, delivering sharp, well-calibrated uncertainties. This work establishes the neural process family as a scalable, probabilistic foundation for real-time transient science in the Rubin era.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper introduces the neural process family, specifically Attentive Neural Processes (NightLANP), for probabilistic, multi-band light-curve reconstruction of astrophysical transients. By meta-learning on simulated transients across 15 classes with realistic Rubin cadences, the model performs amortized inference that is class-agnostic, outperforms Gaussian Processes and neural-network baselines on regression quality, astrophysical feature recovery, and probabilistic calibration, and achieves microsecond-scale interpolation of all bands simultaneously—four to five orders of magnitude faster than the next-best methods.

Significance. If the sim-to-real transfer holds, the approach would supply a scalable, probabilistic foundation for real-time processing of the Rubin alert stream, removing the need for per-curve kernel specification or individual fitting while delivering well-calibrated uncertainties that avoid both the overconfidence of standard neural nets and the underconfidence of GPs.

major comments (2)
  1. [Evaluation section] Evaluation section (and abstract): all reported outperformance metrics, feature-recovery scores, calibration diagnostics, and timing benchmarks are obtained exclusively on held-out simulated light curves; no experiments on actual Rubin or other survey observations are presented. This is load-bearing for the central claim that the meta-learned, class-agnostic model generalizes to real data without domain-specific kernels or per-curve fitting.
  2. [Training and data-generation section] § on model training and data generation: the premise that meta-learning on the chosen suite of 15 simulated transient classes captures the distribution shifts in noise, systematics, cadence irregularities, and morphologies present in real observations is stated but not tested; the transferability argument therefore rests on an unverified assumption.
minor comments (3)
  1. [Abstract] Abstract and introduction: the phrase “even an unoptimized, out-of-the-box Attentive Neural Process” is repeated; a single occurrence with a brief definition of the exact architecture and hyper-parameters used would suffice.
  2. [Introduction] Notation: the distinction between the full neural-process family and the specific Attentive Neural Process variant is not always maintained; consistent use of “ANP” versus “NP family” would improve clarity.
  3. [Figures] Figure captions: several panels lack explicit axis labels or units for the recovered astrophysical features; adding these would aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed report. We address each major comment point by point below, with honest acknowledgment of the simulation-only scope of the current work.

read point-by-point responses

  1. Referee: [Evaluation section] Evaluation section (and abstract): all reported outperformance metrics, feature-recovery scores, calibration diagnostics, and timing benchmarks are obtained exclusively on held-out simulated light curves; no experiments on actual Rubin or other survey observations are presented. This is load-bearing for the central claim that the meta-learned, class-agnostic model generalizes to real data without domain-specific kernels or per-curve fitting.

    Authors: The evaluation is performed exclusively on held-out simulated light curves, as stated throughout the manuscript. Our claims are scoped to outperformance under realistic Rubin cadences in simulation; the abstract refers to 'demonstrating the potential' rather than asserting proven generalization to real observations. We will revise the abstract and add an explicit limitations paragraph in the discussion to clarify the simulation-based nature of all reported metrics and to frame real-data transfer as future work. This makes the scope of the claims precise without altering the presented results. revision: partial

  2. Referee: [Training and data-generation section] § on model training and data generation: the premise that meta-learning on the chosen suite of 15 simulated transient classes captures the distribution shifts in noise, systematics, cadence irregularities, and morphologies present in real observations is stated but not tested; the transferability argument therefore rests on an unverified assumption.

    Authors: We agree that transferability to real observations is an untested assumption. The simulations incorporate realistic cadences, noise models, and transient morphologies, but they cannot fully capture all real-world systematics. In the revised manuscript we will expand the data-generation and discussion sections to state this limitation explicitly and to outline possible future directions such as domain adaptation or targeted real-data fine-tuning. revision: yes

Circularity Check

0 steps flagged

No circularity; evaluation metrics computed independently on held-out simulations against external benchmarks

full rationale

The paper trains an Attentive Neural Process via meta-learning on simulated transients and reports regression quality, astrophysical feature recovery, and probabilistic calibration on separate held-out simulated test sets. These metrics are defined externally (e.g., against ground-truth light curves and compared to GP and NN baselines) rather than reducing to quantities defined by the model's own fitted parameters or by self-citation chains. No equations or claims exhibit self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations; the amortized inference advantage follows directly from the standard neural process architecture applied to the domain. The derivation remains self-contained against the stated benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only; free parameters and axioms cannot be enumerated from the provided text. The central claim rests on the unstated premise that simulated training data distribution matches real observations sufficiently for generalization.

pith-pipeline@v0.9.1-grok · 5849 in / 1255 out tokens · 31662 ms · 2026-07-01T16:01:08.932971+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

109 extracted references · 91 canonical work pages · 16 internal anchors

  1. [1]

    2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems

    Abadi, M., Agarwal, A., Barham, P., et al. 2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. https://www.tensorflow.org/

    2015

  2. [2]

    2023, Annual Review of Astronomy and Astrophysics, 61, 329

    Aigrain, S., & Foreman-Mackey, D. 2023, Annual Review of Astronomy and Astrophysics, 61, 329

    2023

  3. [3]

    Chronos: Learning the Language of Time Series

    Ansari, A. F., Stella, L., Turkmen, C., et al. 2024, Chronos: Learning the Language of Time Series, arXiv, doi: 10.48550/ARXIV.2403.07815 17 https://www.lsstcorporation.org/science-collaborations 18 https://lsst-tvssc.github.io/

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2403.07815 2024

  4. [4]

    2024, Translation Equivariant Transformer Neural Processes, arXiv, doi: 10.48550/arXiv.2406.12409

    Ashman, M., Diaconu, C., Kim, J., et al. 2024, Translation Equivariant Transformer Neural Processes, arXiv, doi: 10.48550/arXiv.2406.12409

    work page doi:10.48550/arxiv.2406.12409 2024

  5. [5]

    C., Kulkarni, S

    Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2019, Publications of the Astronomical Society of the Pacific, 131, 018002, doi: 10.1088/1538-3873/aaecbe

    work page doi:10.1088/1538-3873/aaecbe 2019

  6. [6]

    2025, A Photometric Classifier for Tidal Disruption Events in Rubin LSST, arXiv, doi: 10.48550/arXiv.2509.25902

    Bhardwaj, K., Christov, A., & Karpov, S. 2025, A Photometric Classifier for Tidal Disruption Events in Rubin LSST, arXiv, doi: 10.48550/arXiv.2509.25902

    work page doi:10.48550/arxiv.2509.25902 2025

  7. [7]

    E., & Fabbro, S

    Bi, C., Woods, T. E., & Fabbro, S. 2024, The Astrophysical Journal, 964, 193, doi: 10.3847/1538-4357/ad1b5a 20

    work page doi:10.3847/1538-4357/ad1b5a 2024

  8. [8]

    B., Ivezi´ c, Z., Jones, R

    Bianco, F. B., Ivezi´ c, Z., Jones, R. L., et al. 2021, The Astrophysical Journal Supplement Series, 258, 1, doi: 10.3847/1538-4365/ac3e72

    work page doi:10.3847/1538-4365/ac3e72 2021

  9. [9]

    A., Gagliano, A., Nugent, A., & Hsu, B

    Boesky, A., Villar, V. A., Gagliano, A., Nugent, A., & Hsu, B. 2026, The Astrophysical Journal Supplement Series, 283, 55, doi: 10.3847/1538-4365/ae4350

    work page doi:10.3847/1538-4365/ae4350 2026

  10. [10]

    2026, Astronomy & Astrophysics, 708, A325, doi: 10.1051/0004-6361/202558816

    Bommireddy, H., F¨ orster, F., McMahon, I., et al. 2026, Astronomy & Astrophysics, 708, A325, doi: 10.1051/0004-6361/202558816

    work page doi:10.1051/0004-6361/202558816 2026

  11. [11]

    2019, The Astronomical Journal, 158, 257, doi: 10.3847/1538-3881/ab5182

    Boone, K. 2019, The Astronomical Journal, 158, 257, doi: 10.3847/1538-3881/ab5182

    work page doi:10.3847/1538-3881/ab5182 2019

  12. [12]

    2018, JAX: composable transformations of Python+NumPy programs, 0.3.13

    Bradbury, J., Frostig, R., Hawkins, P., et al. 2018, JAX: composable transformations of Python+NumPy programs, 0.3.13. http://github.com/jax-ml/jax

    2018

  13. [13]

    Rapid and robust simulation-based inference for kilonovae

    Brown, S. M., Bulla, M., Peiris, H. V., et al. 2026, Rapid and robust simulation-based inference for kilonovae, arXiv, doi: 10.48550/arXiv.2605.13983

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.13983 2026

  14. [14]

    P., Requeima, J., Foong, A

    Bruinsma, W. P., Requeima, J., Foong, A. Y. K., Gordon, J., & Turner, R. E. 2021, The Gaussian Neural Process, arXiv, doi: 10.48550/arXiv.2101.03606

    work page doi:10.48550/arxiv.2101.03606 2021

  15. [15]

    P., Markou, S., Requiema, J., et al

    Bruinsma, W. P., Markou, S., Requiema, J., et al. 2023, Autoregressive Conditional Neural Processes, arXiv, doi: 10.48550/arXiv.2303.14468

    work page doi:10.48550/arxiv.2303.14468 2023

  16. [16]

    2024, Astronomy and Astrophysics, 689, A289, doi: 10.1051/0004-6361/202449475

    Cabrera-Vives, G., Moreno-Cartagena, D., Astorga, N., et al. 2024, Astronomy and Astrophysics, 689, A289, doi: 10.1051/0004-6361/202449475

    work page doi:10.1051/0004-6361/202449475 2024

  17. [17]

    B., & Mahabal, A

    Chaini, S., Bianco, F. B., & Mahabal, A. 2025, In Search of the Unknown Unknowns: A Multi-Metric Distance Ensemble for Out of Distribution Anomaly Detection in Astronomical Surveys, arXiv, doi: 10.48550/arXiv.2510.23702

    work page doi:10.48550/arxiv.2510.23702 2025

  18. [18]

    Chaini, S., Mahabal, A., Kembhavi, A., & Bianco, F. B. 2024, Astronomy and Computing, 48, 100850, doi: 10.1016/j.ascom.2024.100850

    work page doi:10.1016/j.ascom.2024.100850 2024

  19. [19]

    The Pan-STARRS1 Surveys

    Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2019, The Pan-STARRS1 Surveys. https://arxiv.org/abs/1612.05560

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  20. [20]

    2016, in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 785–794

    Chen, T., & Guestrin, C. 2016, in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 785–794

    2016

  21. [21]

    2015, Keras, https://keras.io

    Chollet, F., et al. 2015, Keras, https://keras.io

    2015

  22. [22]

    1997, The Astrophysical Journal, 483, 675

    Clocchiatti, A., Wheeler, J., Phillips, M., et al. 1997, The Astrophysical Journal, 483, 675

    1997

  23. [23]

    2024, Serbian Astronomical Journal, 17, doi: 10.2298/SAJ2408017C da Costa-Luis, C

    Cvorovic-Hajdinjak, I. 2024, Serbian Astronomical Journal, 17, doi: 10.2298/SAJ2408017C da Costa-Luis, C. O. 2019, Journal of Open Source Software, 4, 1277, doi: 10.21105/joss.01277

    work page doi:10.2298/saj2408017c 2024

  24. [24]

    LightCurveLynx: Forward Modeling of Time-Domain Surveys with Application to ZTF SN Ia DR2

    Dai, M., Kubica, J., Malanchev, K., et al. 2026, LightCurveLynx: Forward Modeling of Time-Domain Surveys with Application to ZTF SN Ia DR2, arXiv, doi: 10.48550/arXiv.2604.07134

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2604.07134 2026

  25. [25]

    A decoder-only foundation model for time-series forecasting

    Das, A., Kong, W., Sen, R., & Zhou, Y. 2023, A decoder-only foundation model for time-series forecasting, arXiv, doi: 10.48550/ARXIV.2310.10688

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2310.10688 2023

  26. [26]

    M., Tzanidakis, A., et al

    De, K., Kasliwal, M. M., Tzanidakis, A., et al. 2020, The Astrophysical Journal, 905, 58, doi: 10.3847/1538-4357/abb45c

    work page doi:10.3847/1538-4357/abb45c 2020

  27. [27]

    2023, Astronomy and Astrophysics, 677, A16, doi: 10.1051/0004-6361/202245189

    Demianenko, M., Malanchev, K., Samorodova, E., et al. 2023, Astronomy and Astrophysics, 677, A16, doi: 10.1051/0004-6361/202245189

    work page doi:10.1051/0004-6361/202245189 2023

  28. [28]

    2025, Astromer 2, doi: 10.48550/arXiv.2502.02717 —

    Donoso-Oliva, C., Becker, I., Protopapas, P., et al. 2025, Astromer 2, doi: 10.48550/arXiv.2502.02717 —. 2023, Astronomy and Astrophysics, 670, A54, doi: 10.1051/0004-6361/202243928

    work page doi:10.48550/arxiv.2502.02717 2025

  29. [29]

    Dubois, Y., Gordon, J., & Foong, A. Y. 2020, Neural Process Family, http://yanndubs.github.io/Neural-Process-Family/

    2020

  30. [30]

    2017, The Astronomical Journal, 154, 220, doi: 10.3847/1538-3881/aa9332

    Foreman-Mackey, D., Agol, E., Ambikasaran, S., & Angus, R. 2017, The Astronomical Journal, 154, 220, doi: 10.3847/1538-3881/aa9332

    work page internal anchor Pith review doi:10.3847/1538-3881/aa9332 2017

  31. [31]

    2020, The Astrophysical Journal, 895, 32

    Fremling, C., Miller, A., Sharma, Y., et al. 2020, The Astrophysical Journal, 895, 32

    2020

  32. [32]

    Wilson, A. G. 2021, GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration, arXiv, doi: 10.48550/arXiv.1809.11165

    work page doi:10.48550/arxiv.1809.11165 2021

  33. [33]

    Neural Processes

    Garnelo, M., Schwarz, J., Rosenbaum, D., et al. 2018a, Neural Processes, arXiv, doi: 10.48550/arXiv.1807.01622

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.01622

  34. [34]

    Conditional Neural Processes

    Garnelo, M., Rosenbaum, D., Maddison, C. J., et al. 2018b, Conditional Neural Processes, arXiv, doi: 10.48550/arXiv.1807.01613

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.01613

  35. [35]

    TimeGPT-1,

    Garza, A., Challu, C., & Mergenthaler-Canseco, M. 2023, TimeGPT-1, arXiv, doi: 10.48550/ARXIV.2310.03589

    work page doi:10.48550/arxiv.2310.03589 2023

  36. [36]

    D., Angerhausen, D., Konrad, B

    Gebhard, T. D., Angerhausen, D., Konrad, B. S., et al. 2024, Astronomy & Astrophysics, 681, A3, doi: 10.1051/0004-6361/202346390

    work page doi:10.1051/0004-6361/202346390 2024

  37. [37]

    Gneiting, T., Balabdaoui, F., & Raftery, A. E. 2007, Journal of the Royal Statistical Society Series B: Statistical Methodology, 69, 243, doi: 10.1111/j.1467-9868.2007.00587.x

    work page doi:10.1111/j.1467-9868.2007.00587.x 2007

  38. [38]

    Gneiting, T., & Raftery, A. E. 2007, Journal of the American Statistical Association, 102, 359, doi: 10.1198/016214506000001437

    work page doi:10.1198/016214506000001437 2007

  39. [39]

    P., Foong, A

    Gordon, J., Bruinsma, W. P., Foong, A. Y. K., et al. 2020, Convolutional Conditional Neural Processes, arXiv, doi: 10.48550/arXiv.1910.13556 21

    work page doi:10.48550/arxiv.1910.13556 2020

  40. [40]

    A., et al

    Guillochon, J., Nicholl, M., Villar, V. A., et al. 2018, The Astrophysical Journal Supplement Series, 236, 6, doi: 10.3847/1538-4365/aab761

    work page doi:10.3847/1538-4365/aab761 2018

  41. [41]

    Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. 2017, in International conference on machine learning, PMLR, 1321–1330, doi: 10.48550/arXiv.1706.04599

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1706.04599 2017

  42. [42]

    P., AlSayyad, Y., Bechtol, K., et al

    Guy, L. P., AlSayyad, Y., Bechtol, K., et al. 2026, Rubin Observatory Plans for an Early Science Program, Technical Note RTN-011, NSF-DOE Vera C. Rubin Observatory, doi: 10.71929/rubin/2584021

    work page doi:10.71929/rubin/2584021 2026

  43. [43]

    A., & Rosenbaum, D

    Hamad, H. A., & Rosenbaum, D. 2025, Flow Matching Neural Processes, arXiv, doi: 10.48550/arXiv.2512.23853

    work page doi:10.48550/arxiv.2512.23853 2025

  44. [44]

    M., Bianco, F

    Hambleton, K. M., Bianco, F. B., Street, R., et al. 2022, Rubin Observatory LSST Transients and Variable Stars Roadmap. https://arxiv.org/abs/2208.04499

    work page arXiv 2022

  45. [45]

    Review on the Observed and Physical Properties of Core Collapse Supernovae

    Hamuy, M. 2003, in 2003 Aspen Summer Workshop on the Nuclear Physics of Core Collapse Supernovae. https://arxiv.org/abs/astro-ph/0301006

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  46. [46]

    R., Millman, K

    Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2 Hloˇ zek, R., Malz, A. I., Ponder, K. A., et al. 2023, The Astrophysical Journal Supplement Series, 267, 25, doi: 10.3847/1538-4365/accd6a

    work page doi:10.1038/s41586-020-2649-2 2020

  47. [47]

    Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55

    work page doi:10.1109/mcse.2007.55 2007

  48. [48]

    Hurlbert, S. H. 1984, Ecological Monographs, 54, 187, doi: 10.2307/1942661

    work page doi:10.2307/1942661 1984

  49. [49]

    2025, PERFECT PLASTICC SIM, Zenodo, doi: 10.5281/ZENODO.17392460 Ivezi´ c, Z., Kahn, S

    Ishida, E., & Jones, D. 2025, PERFECT PLASTICC SIM, Zenodo, doi: 10.5281/ZENODO.17392460 Ivezi´ c, Z., Kahn, S. M., Tyson, J. A., et al. 2019, The Astrophysical Journal, 873, 111, doi: 10.3847/1538-4357/ab042c

    work page doi:10.5281/zenodo.17392460 2025

  50. [50]

    R., Arredondo, J., Valenzuela, C., et al

    Jainaga, I. R., Arredondo, J., Valenzuela, C., et al. 2021, alercebroker/lc classifier: Release 1.2.3-P, Zenodo, doi: 10.5281/ZENODO.5275453

    work page doi:10.5281/zenodo.5275453 2021

  51. [51]

    B., Ili´ c, D., et al

    Jankov, I., Kovaˇ cevi´ c, A. B., Ili´ c, D., et al. 2022, Astronomische Nachrichten, 343, e210090, doi: 10.1002/asna.20210090

    work page doi:10.1002/asna.20210090 2022

  52. [52]

    E., & Yao, L

    Jha, S., Gong, D., Wang, X., Turner, R. E., & Yao, L. 2023, The Neural Process Family: Survey, Applications and Perspectives, arXiv, doi: 10.48550/arXiv.2209.00517

    work page doi:10.48550/arxiv.2209.00517 2023

  53. [53]

    M., et al

    Karchev, K., Grayling, M., Boyd, B. M., et al. 2024, Monthly Notices of the Royal Astronomical Society, 530, 3881, doi: 10.1093/mnras/stae995

    work page doi:10.1093/mnras/stae995 2024

  54. [54]

    2017, Monthly Notices of the Royal Astronomical Society, 467, 3299

    Kashi, A., & Soker, N. 2017, Monthly Notices of the Royal Astronomical Society, 467, 3299

    2017

  55. [55]

    M., Kulkarni, S., Gal-Yam, A., et al

    Kasliwal, M. M., Kulkarni, S., Gal-Yam, A., et al. 2012, The Astrophysical Journal, 755, 161

    2012

  56. [56]

    Kennamer, N., Ishida, E. E. O., Gonzalez-Gaitan, S., et al. 2020, Active learning with RESSPECT: Resource allocation for extragalactic astronomical transients, arXiv, doi: 10.48550/arXiv.2010.05941

    work page doi:10.48550/arxiv.2010.05941 2020

  57. [57]

    P., Cinabro, D., et al

    Kessler, R., Bernstein, J. P., Cinabro, D., et al. 2009, Publications of the Astronomical Society of the Pacific, 121, 1028, doi: 10.1086/605984

    work page doi:10.1086/605984 2009

  58. [58]

    B., Modjaz, M., et al

    Khakpash, S., Bianco, F. B., Modjaz, M., et al. 2024, The Astrophysical Journal Supplement Series, 275, 37, doi: 10.3847/1538-4365/ad7eaa

    work page doi:10.3847/1538-4365/ad7eaa 2024

  59. [59]

    Attentive Neural Processes

    Kim, H., Mnih, A., Schwarz, J., et al. 2019, Attentive Neural Processes, arXiv, doi: 10.48550/arXiv.1901.05761

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.05761 2019

  60. [60]

    2016, in Positioning and Power in Academic Publishing: Players, Agents and Agendas, ed

    Kluyver, T., Ragan-Kelley, B., P´ erez, F., et al. 2016, in Positioning and Power in Academic Publishing: Players, Agents and Agendas, ed. F. Loizides & B. Schmidt, IOS Press, 87 – 90

    2016

  61. [61]

    B., Ilic, D., Popovic, L

    Kovacevic, A. B., Ilic, D., Popovic, L. C., et al. 2023, Deep learning of quasar lightcurves in the LSST era, arXiv, doi: 10.48550/arXiv.2306.09357

    work page doi:10.48550/arxiv.2306.09357 2023

  62. [62]

    Li, W., Chen, H.-Y., Rehemtulla, N., et al. 2026, StarEmbed: Benchmarking Time Series Foundation Models on Astronomical Observations of Variable Stars, arXiv, doi: 10.48550/arXiv.2510.06200 LSST Dark Energy Science Collaboration, Aubourg, E.,

    work page doi:10.48550/arxiv.2510.06200 2026

  63. [63]

    2026, Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration, Tech

    Avestruz, C., et al. 2026, Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration, Tech. rep., doi: 10.48550/arXiv.2601.14235

    work page doi:10.48550/arxiv.2601.14235 2026

  64. [64]

    H., Gunn, J

    Lupton, R. H., Gunn, J. E., & Szalay, A. S. 1999, The Astronomical Journal, 118, 1406, doi: 10.1086/301004

    work page doi:10.1086/301004 1999

  65. [65]

    2026, RAS Techniques and Instruments, 5, rzag019, doi: 10.1093/rasti/rzag019

    Magill, D., Nicholl, M., Anilkumar, V., et al. 2026, RAS Techniques and Instruments, 5, rzag019, doi: 10.1093/rasti/rzag019

    work page doi:10.1093/rasti/rzag019 2026

  66. [66]

    2017, in 2017 IEEE Symposium Series on Computational Intelligence (SSCI) (Honolulu, HI: IEEE), 1–8, doi: 10.1109/SSCI.2017.8280984

    Mahabal, A., Sheth, K., Gieseke, F., et al. 2017, in 2017 IEEE Symposium Series on Computational Intelligence (SSCI) (Honolulu, HI: IEEE), 1–8, doi: 10.1109/SSCI.2017.8280984

    work page doi:10.1109/ssci.2017.8280984 2017

  67. [67]

    2021, Astrophysics Source Code Library, ascl:2107.001

    Malanchev, K. 2021, Astrophysics Source Code Library, ascl:2107.001. https://ui.adsabs.harvard.edu/abs/2021ascl.soft07001M

    2021

  68. [68]

    L., Pruzhinskaya, M

    Malanchev, K. L., Pruzhinskaya, M. V., Korolev, V. S., et al. 2021, Monthly Notices of the Royal Astronomical Society, 502, 5147, doi: 10.1093/mnras/stab316

    work page doi:10.1093/mnras/stab316 2021

  69. [69]

    Mortimer, P., Diaconu, C., Rochussen, T., Mlodozeniec, B., & Turner, R. E. 2026, Incremental Transformer Neural Processes, arXiv, doi: 10.48550/arXiv.2602.18955

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2602.18955 2026

  70. [70]

    S., Biswas, R., & Hloˇ zek, R

    Muthukrishna, D., Narayan, G., Mandel, K. S., Biswas, R., & Hloˇ zek, R. 2019, Publications of the Astronomical Society of the Pacific, 131, 118002, doi: 10.1088/1538-3873/ab1609 22 M¨ oller, A., & de Boissi` ere, T. 2020, Monthly Notices of the Royal Astronomical Society, 491, 4277, doi: 10.1093/mnras/stz3312

    work page doi:10.1088/1538-3873/ab1609 2019

  71. [71]

    2023, The Extended LSST Astronomical Time-series Classification Challenge (ELAsTiCC)

    Narayan, G., & ELAsTiCC Team. 2023, The Extended LSST Astronomical Time-series Classification Challenge (ELAsTiCC). https://ui.adsabs.harvard.edu/abs/2023AAS...24111701N O’Mullane, W., Dubois, R., Butler, M., & Lim, K. T. 2023, DM sizing model and cost plan for construction and operations, Project Science Technical Note DMTN-135, NSF-DOE Vera C. Rubin Obs...

    work page doi:10.5281/zenodo.3509134 2023

  72. [72]

    W., Villar, A., Li, Y., et al

    Park, J. W., Villar, A., Li, Y., et al. 2021, Inferring Black Hole Properties from Astronomical Multivariate Time Series with Bayesian Attentive Neural Processes, arXiv, doi: 10.48550/arXiv.2106.01450

    work page doi:10.48550/arxiv.2106.01450 2021

  73. [73]

    2019, PyTorch: an imperative style, high-performance deep learning library (Red Hook, NY, USA: Curran Associates Inc.)

    Paszke, A., Gross, S., Massa, F., et al. 2019, PyTorch: an imperative style, high-performance deep learning library (Red Hook, NY, USA: Curran Associates Inc.)

    2019

  74. [74]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    2011

  75. [75]

    GOPREAUX I: Open-source Code and Data to Model Multi-wavelength Emission of Extragalactic Transients using Gaussian Processes

    Pellegrino, C., Pritchard, T. A., Modjaz, M., et al. 2026, arXiv e-prints, arXiv:2604.03372, doi: 10.48550/arXiv.2604.03372

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2604.03372 2026

  76. [76]

    2024, Physical Review Research, 6, 033078, doi: 10.1103/PhysRevResearch.6.033078

    Peng, Y., Risti´ c, M., Kedia, A., et al. 2024, Physical Review Research, 6, 033078, doi: 10.1103/PhysRevResearch.6.033078

    work page doi:10.1103/physrevresearch.6.033078 2024

  77. [77]

    A., Fremling, C., Sollerman, J., et al

    Perley, D. A., Fremling, C., Sollerman, J., et al. 2020, The Astrophysical Journal, 904, 35

    2020

  78. [78]

    2025, Design-marginal calibration of Gaussian process predictive distributions: Bayesian and conformal approaches, Hyper Articles en Ligne

    Pion, A., & Vazquez, E. 2025, Design-marginal calibration of Gaussian process predictive distributions: Bayesian and conformal approaches, Hyper Articles en Ligne. https://centralesupelec.hal.science/hal-05400282 PLAsTiCC Modelers. 2022, Libraries & Recommended Citations for using PLAsTiCC Models, Zenodo, doi: 10.5281/ZENODO.2612895

    work page doi:10.5281/zenodo.2612895 2025

  79. [79]

    2022, The Astronomical Journal, 163, 57, doi: 10.3847/1538-3881/ac39a1

    Qu, H., & Sako, M. 2022, The Astronomical Journal, 163, 57, doi: 10.3847/1538-3881/ac39a1

    work page doi:10.3847/1538-3881/ac39a1 2022

  80. [80]

    E., & Williams, C

    Rasmussen, C. E., & Williams, C. K. I. 2008, Gaussian processes for machine learning, 3rd edn., Adaptive computation and machine learning (Cambridge, Mass.: MIT Press)

    2008

Showing first 80 references.