pith. sign in

arxiv: 2605.18959 · v1 · pith:QEZ5ZFS5new · submitted 2026-05-18 · 🌌 astro-ph.IM · astro-ph.CO· astro-ph.EP· astro-ph.GA· cs.LG

Hyrax: An Extensible Framework for Rapid ML Experimentation and Unsupervised Discovery in the Era of Rubin, Roman, and Euclid

Aritra Ghosh , Drew Oldag , Michael Tauraso , Andrew J. Connolly , Peter Ferguson , Derek Jones , Gourav Khullar , Argyro Sasli
show 24 more authors
Samarth Venkatesh Gracia Wang Maxine West Dylan Berry Neven Caplar Colin Orion Chandler Tanawan Chatchadanoraset Michael W. Coughlin Melissa DeLucchi Alexandra Junell Diego Miura Felipe Fontinele Nunes Wilson Beebe Doug Branton Sandro Campos Liam Cunningham Mi Dai Jeremy Kubica Konstantin Malanchev Rachel Mandelbaum Sean McGuire Imad Pasha Dan S. Taranu Tianqing Zhang
This is my paper

Pith reviewed 2026-05-20 07:26 UTC · model grok-4.3

classification 🌌 astro-ph.IM astro-ph.COastro-ph.EPastro-ph.GAcs.LG
keywords machine learning frameworkastronomical surveysunsupervised discoveryRubin LSSTlatent space explorationvector databasesgravitational lensestransient classification
0
0 comments X p. Extension
pith:QEZ5ZFS5 Add to your LaTeX paper What is a Pith Number? 
\usepackage{pith}
\pithnumber{QEZ5ZFS5}

Prints a linked pith:QEZ5ZFS5 badge after your title and writes the identifier into PDF metadata. Compiles on arXiv with no extra files. Learn more

The pith

Hyrax framework manages the full machine learning pipeline for large astronomical surveys to support unsupervised discovery.

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

The paper presents Hyrax as an open-source modular Python framework designed for astronomical machine learning tasks at the scale of upcoming surveys like Rubin, Roman, and Euclid. It integrates data handling, training, inference, vector databases for similarity search, and tools for exploring latent spaces to enable rapid experiments and discovery without extensive custom infrastructure. Demonstrations include processing hundreds of thousands of galaxies from Rubin data to identify new merger candidates and low-surface-brightness objects purely through unsupervised methods, along with other applications in clustering lenses, classifying transients, and detecting solar system objects. A sympathetic reader would care because this setup aims to move the main effort from building data systems to designing and iterating on scientific models for massive datasets.

Core claim

Hyrax is an extensible open-source GPU-enabled Python framework that supports the full ML lifecycle in astronomy through modular components for data acquisition, multimodal datasets, integrated vector databases for similarity search, and interactive latent-space exploration, as demonstrated by applications such as unsupervised representation learning on approximately 400,000 Rubin DP1 galaxies that identifies new merger and low-surface-brightness candidates absent from existing catalogs while isolating artifacts without labeled data.

What carries the argument

Hyrax's modular architecture integrating vector databases and interactive two- and three-dimensional latent-space exploration tools that carry out similarity searches and unsupervised discovery on survey-scale data.

If this is right

  • Unsupervised methods can surface previously uncatalogued galaxy mergers and low-surface-brightness objects in Rubin DP1 data.
  • Hybrid density-based clustering can flag cluster-scale gravitational lens candidates directly from imaging.
  • Multimodal inputs enable early-time classification of transients combining light curves, spectra, images, and metadata.
  • Supervised filtering reduces false positives in searches for distant solar system objects using shift-and-stack techniques.
  • Synthetic source injection supports detection of semi-resolved dwarf galaxies in high-resolution imaging surveys.

Where Pith is reading between the lines

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

  • Smaller research teams could perform systematic discovery work on next-generation survey data without building large custom software stacks.
  • The latent-space tools might generalize to other domains where unlabeled high-volume imaging data requires rapid candidate identification.
  • Adoption could standardize experiment comparison across astronomy ML projects by providing built-in logging and visualization.

Load-bearing premise

That the modular design with built-in tools for vector search and latent exploration will shift the main effort from infrastructure setup to scientific model design for groups handling full volumes of data from Rubin, Roman, and Euclid.

What would settle it

A side-by-side timing comparison of setting up and running an unsupervised representation learning task on a 400,000-object galaxy catalog using Hyrax versus a standard custom pipeline without the framework's integrated components.

Figures

Figures reproduced from arXiv: 2605.18959 by Alexandra Junell, Andrew J. Connolly, Argyro Sasli, Aritra Ghosh, Colin Orion Chandler, Dan S. Taranu, Derek Jones, Diego Miura, Doug Branton, Drew Oldag, Dylan Berry, Felipe Fontinele Nunes, Gourav Khullar, Gracia Wang, Imad Pasha, Jeremy Kubica, Konstantin Malanchev, Liam Cunningham, Maxine West, Melissa DeLucchi, Michael Tauraso, Michael W. Coughlin, Mi Dai, Neven Caplar, Peter Ferguson, Rachel Mandelbaum, Samarth Venkatesh, Sandro Campos, Sean McGuire, Tanawan Chatchadanoraset, Tianqing Zhang, Wilson Beebe.

Figure 1
Figure 1. Figure 1: Survey grasp—defined as survey area scaled by angular resolution—is plotted against flux sensitivity for a selection of representative astronomical imaging surveys. Legacy surveys are shown as blue circles, while upcoming wide-fast-deep surveys (e.g., Rubin-LSST, Euclid, Roman, and CSST) are shown as red squares. The diagonal ar￾row indicates the direction of increasing information con￾tent, which scales w… view at source ↗
Figure 2
Figure 2. Figure 2: Annual number of peer-reviewed astronomy jour￾nal articles indexed in the NASA Astrophysics Data System (ADS) from 2010–2024 containing the indicated keywords in either the title or abstract. The steep, sustained rise in publications referencing “machine learning” and “neural net￾work” over the past decade reflects the rapid adoption of these methods across the astronomical community, in con￾trast to the r… view at source ↗
Figure 3
Figure 3. Figure 3: The multitude of different components involved in a typical ML project in astronomy. While the ML model code itself is only one piece of the workflow, successful projects require substantial surrounding infrastructure for tasks such as data acquisition, efficient data loading, run monitoring, and interactive visualization of latent spaces. Hyrax provides robust, astronomy-specific implementations for all o… view at source ↗
Figure 4
Figure 4. Figure 4: Schematic of the primary modules within Hyrax. (1) The workflow begins with hyrax.download() for parallelized data retrieval from supported surveys or custom user-provided sources. (2) Retrieved data are organized into flexible, astron￾omy-tailored ML datasets. (3) The training module (hyrax.train()) offers intelligent pre-caching, live monitoring, logging, and multi-GPU support, via e.g., PyTorch, MLflow,… view at source ↗
Figure 5
Figure 5. Figure 5: Training ML models on survey-scale imaging data requires efficient access to potentially millions of image cutouts. This block diagram shows how Hyrax’s LSSTDataset class addresses this: survey patches are retrieved via the Rubin Butler and cached at the patch level on local storage, preventing repeated queries for the same sky region. During training, image cutouts are preemptively generated on the fly fr… view at source ↗
Figure 6
Figure 6. Figure 6: Diagram of the primary components that control the flow of data during training and inference in Hyrax. Data are read by an appropriate DatasetClass that exposes components of the data to DataProvider. The DataProvider prepares a combined dataset based on the user data request configuration. The ML engine will request data samples for training or inference, which are then furnished by DataProvider. The Dat… view at source ↗
Figure 7
Figure 7. Figure 7: Model comparison and experiment tracking in Hyrax via its native MLflow integration. The static version of this figure shows the MLflow interface populated automatically during Hyrax training runs, displaying loss curves, evaluation metrics, and hyperparameters for multiple experiments executed under a common configuration framework. This enables systematic and easy comparison of models across different da… view at source ↗
Figure 8
Figure 8. Figure 8: Similarity search examples performed on Ru￾bin DP1 galaxies using an unsupervised ML model within the Hyrax framework (see §3.1 for details). The left col￾umn shows query images, while the right columns show nearest neighbors retrieved via the search by vector() method. In the top row, a query galaxy exhibiting extended low–surface-brightness features retrieves neigh￾bors with similarly diffuse outer light… view at source ↗
Figure 9
Figure 9. Figure 9: Two-dimensional latent-space visualization produced using the Hyrax 2D Latent Space Explorer. The animated version of this figure (available in the online version and at this link) demonstrates interactive exploration of the latent space, allowing users to color objects by catalog-level properties and to select regions of the embedding space for further inspection. The static snapshot also shows the hyrax.… view at source ↗
Figure 10
Figure 10. Figure 10: Hyrax’s browser-based 3D Latent Space Explorer allows users to explore three-dimensional representations of learned latent spaces interactively. The static image shown here is a snapshot from a video figure, and the video is available in the online version of the article and at this link. Readers can also try out a live interactive demonstration at this link. The interface allows users to rotate, pan, and… view at source ↗
Figure 11
Figure 11. Figure 11: Two latent-space representations for the i < 20 Rubin DP1 sources in ECDFS. (Left:) Filter-specific artifacts/defects naturally segregate into a distinct island. (Right:) Removing these and retraining reveals a second generation of outliers, now comprising imaging artifacts such as diffraction spikes and satellite trails/cosmic-ray hits. Each removal round progressively drives the sample toward a cleaner … view at source ↗
Figure 12
Figure 12. Figure 12: UMAP projections of latent spaces learned by three unsupervised models with varying hyperparameters, applied to 20 ≤ i < 22 ECDFS DP1 galaxies. Overlaid blue squares mark mergers identified in the Euclid Quick Release 1 visual morphology catalog ( Euclid Collaboration et al. 2025a). The spatial clustering of the mergers varies non-monotonically with the validation loss obtained for each model (shown above… view at source ↗
Figure 13
Figure 13. Figure 13: (Right:) The UMAP that corresponds to the strongest clustering of merger candidates for the 20 ≤ i < 22 Rubin DP1 ECDFS sample is shown here, with Euclid merger catalog cross-matches denoted by black squares. We explore the region denoted by the black ellipse using Hyrax’s 2D Latent Space Explorer. (Left:) The top row shows example objects from this region: two randomly selected Euclid-catalog mergers (bl… view at source ↗
Figure 14
Figure 14. Figure 14: (Left:) A UMAP projection of the best-performing latent-space representation for the i < 20 Rubin DP1 ECDFS sample is shown here, with cross-matched DES LSB catalog objects marked as cyan diamonds. The cross-matched objects are compactly clustered within the cyan ellipse, which we investigate further using Hyrax’s 2D latent space explorer. (Right:) The top row shows one randomly selected DES LSB catalog m… view at source ↗
Figure 15
Figure 15. Figure 15: Example of our hybrid lens and arc finding approach. (Left and Center) UMAP and RA+Dec overdensities corresponding to groups and clusters in ECDFS, one of the fields sampled by Rubin DP1 observations. (Right) Post visual inspection, we isolate two lensed arc candidates (marked with solid and dotted lines) in the pan-chromatic Rubin gri and Euclid VIS imaging (see inset). 3.3. Multimodal Learning for ZTF E… view at source ↗
Figure 16
Figure 16. Figure 16: Normalized confusion matrices of AppleCiDEr using (a) three and (b) four modalities. The classification of TDEs is more challenging due to the rarity of this class and the limited number of training samples. The comparison highlights the importance of multimodal training. Hyrax abstracts the surrounding ML infrastructure, allowing complex multimodal architectures to be de￾ployed. The comparatively lower p… view at source ↗
Figure 17
Figure 17. Figure 17: Multimodal visualization of two TDEs. Panel (a) shows a TDE predicted as SN II. Panel (b) shows a correctly classified TDE where stronger spectral and photometric transient signatures are present. The comparison illustrates the chal￾lenges of early-time multimodal classification, where incomplete photometric evolution and limited spectral information can lead to confusion between different classes of opti… view at source ↗
Figure 18
Figure 18. Figure 18: Observations of an inserted synthetic object. In each individual frame, the object is below the detection threshold, but in the coaddition centered on the object’s trajectory, it is clearly visible. poorly constrained, will be essential for testing these and other formation hypotheses. KBMOD (Kernel-Based Moving Object Detection) is a digital tracking or shift-and-stack package that detects moving objects… view at source ↗
Figure 19
Figure 19. Figure 19: Coadded KBMOD result stamps, where the top row represents “true” positive results (i.e., plausible moving objects) and the bottom row represents “false” positive re￾sults (caused by imaging/differencing artifacts or unmasked static objects like galaxies). In its initial implementation, we built a custom Ten￾sorFlow binary classifier distinguishing true and false positives. We recently adopted Hyrax as our… view at source ↗
Figure 20
Figure 20. Figure 20: Left: Example gri-band coadd stamp (50” wide) of HSC SSP data processed using the Rubin Science Pipelines with no source injection. Middle: Same as left image but semi-resolved galaxies at D⊙ ≃ 3.5 Mpc have been injected (circled in white). Right: Results showing detection sensitivity as a function of galaxy size (rh) and brightness (MV ) using a supervised learning workflow with Hyrax and the Rubin Scien… view at source ↗
read the original abstract

The NSF-DOE Vera C. Rubin Observatory, Roman Space Telescope, Euclid, and other next-generation surveys will deliver imaging, spectroscopic, and time-domain data at scales that increasingly shift the bottleneck in astronomical machine learning (ML) projects from model design to infrastructure. We present Hyrax, an open-source, modular, GPU-enabled Python framework that supports the full ML lifecycle in astronomy: from data acquisition and training to inference and experiment comparison, with capabilities including multimodal dataset support, integrated vector databases for similarity search, and interactive two- and three-dimensional latent-space exploration for unsupervised discovery. We demonstrate Hyrax's versatility through five representative applications on real survey data: (i) unsupervised representation learning on $\sim 4\times10^5$ Rubin Legacy Survey of Space and Time (LSST) Data Preview 1 (DP1) galaxies, surfacing new merger and low-surface-brightness candidates missing from reference Euclid and Dark Energy Survey catalogs, while also isolating imaging artifacts -- all without labeled training data; (ii) hybrid density-based clustering for identifying cluster-scale gravitational lens candidates in DP1 data; (iii) multimodal early-time transient classification in the Zwicky Transient Facility leveraging light curves, spectra, images, and metadata; (iv) supervised false-positive filtering in shift-and-stack searches for distant solar system objects in the Dark Energy Camera Ecliptic Exploration Project survey; and (v) supervised detection of semi-resolved dwarf galaxies in Hyper Suprime-Cam and LSST-like imaging using synthetic source injection. Together, these results demonstrate that Hyrax provides astronomy-specific ML infrastructure that enables systematic discovery and rapid methodological iteration across next-generation astronomical surveys.

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

1 major / 2 minor

Summary. The manuscript introduces Hyrax, an open-source modular GPU-enabled Python framework for the full machine-learning lifecycle in astronomy. It supports multimodal datasets, integrated vector databases for similarity search, and interactive latent-space exploration. The central claim is that this infrastructure shifts the bottleneck from model design to systematic discovery and rapid iteration for next-generation surveys. The paper demonstrates the framework via five applications on real data: (i) unsupervised representation learning on ~4e5 Rubin DP1 galaxies that identifies new merger and low-surface-brightness candidates absent from Euclid/DES catalogs while flagging artifacts; (ii) hybrid density-based clustering for cluster-scale gravitational lens candidates in DP1; (iii) multimodal early-time transient classification in ZTF using light curves, spectra, images, and metadata; (iv) supervised false-positive filtering for shift-and-stack solar-system-object searches in DECaLS; and (v) supervised dwarf-galaxy detection in HSC and LSST-like imaging via synthetic source injection.

Significance. If the modular design, vector-DB integration, and latent-space tools function at full survey scale, Hyrax could meaningfully reduce infrastructure overhead for typical research groups working with Rubin, Roman, and Euclid data volumes. The open-source release, use of real survey data, and focus on unsupervised discovery are concrete strengths. However, all five demonstrations operate on DP1-scale or smaller catalogs; without scaling benchmarks the claimed infrastructure advantage for petabyte-scale surveys remains unverified.

major comments (1)
  1. [Section 5] Section 5 (Applications): All five demonstrations use DP1-scale data (~4×10^5 galaxies) or smaller ZTF/DECaLS/HSC subsets. No scaling benchmarks, distributed-training results, memory profiles, or out-of-core performance figures are reported for the 10–100× larger catalogs and petabyte imaging expected from full LSST, Roman, or Euclid. This directly bears on the central claim that Hyrax shifts the infrastructure bottleneck for next-generation surveys.
minor comments (2)
  1. [Abstract] The abstract and introduction could more explicitly distinguish between framework capabilities that are already implemented versus those planned for future releases.
  2. Figure captions for the latent-space visualizations should state the exact dimensionality reduction method and any hyper-parameters used.

Simulated Author's Rebuttal

1 responses · 1 unresolved

We thank the referee for their constructive review and for recognizing the strengths of the open-source release, real-data demonstrations, and focus on unsupervised discovery. We address the major comment below and have revised the manuscript to better contextualize the scope of our claims regarding next-generation survey infrastructure.

read point-by-point responses

  1. Referee: [Section 5] Section 5 (Applications): All five demonstrations use DP1-scale data (~4×10^5 galaxies) or smaller ZTF/DECaLS/HSC subsets. No scaling benchmarks, distributed-training results, memory profiles, or out-of-core performance figures are reported for the 10–100× larger catalogs and petabyte imaging expected from full LSST, Roman, or Euclid. This directly bears on the central claim that Hyrax shifts the infrastructure bottleneck for next-generation surveys.

    Authors: We agree that the lack of explicit scaling benchmarks, distributed-training results, memory profiles, and out-of-core performance figures for 10–100× larger catalogs weakens the direct support for our central claim about shifting infrastructure bottlenecks at full LSST, Roman, or Euclid scales. The five applications were selected to illustrate end-to-end workflows for unsupervised discovery and multimodal analysis using publicly available real survey data at computationally accessible volumes. In the revised manuscript we have added a dedicated subsection (5.6) on scalability considerations. This subsection describes the modular architecture built on PyTorch for GPU acceleration, integration with vector databases (FAISS/Milvus) that support efficient similarity search without full in-memory dataset loading, and data-loading utilities compatible with out-of-core and distributed frameworks such as Dask. We report basic peak-memory and runtime profiles measured on the DP1-scale experiments and provide order-of-magnitude extrapolations assuming linear scaling for embedding generation and indexing steps. We have also tempered the language in the abstract and introduction to emphasize that the current results validate core functionality and discovery utility on representative subsets, while the design is intended to extend to full-survey volumes. Full end-to-end benchmarks on petabyte-scale imaging and distributed training across multiple nodes remain outside the scope of this initial methods paper and would require dedicated large-scale computing allocations. revision: partial

standing simulated objections not resolved
  • Comprehensive distributed-training benchmarks, detailed memory profiles, and out-of-core performance measurements on 10–100× larger catalogs or petabyte-scale imaging from full LSST, Roman, or Euclid

Circularity Check

0 steps flagged

No circularity: framework description and empirical demos are self-contained

full rationale

The paper presents an open-source software framework (Hyrax) together with five empirical applications on real survey data (DP1 galaxies, ZTF transients, DECaLS/DEEP, HSC). No mathematical derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described content. The central claim—that the modular design enables systematic discovery—is supported directly by the listed use cases rather than by any reduction to prior inputs or self-referential definitions. This is the expected non-finding for an infrastructure/tool paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

As a software framework paper rather than a theoretical derivation, the work introduces no free parameters, mathematical axioms, or invented physical entities; all claims rest on the utility of the implemented tools and the representativeness of the five demonstrations.

pith-pipeline@v0.9.0 · 5988 in / 1194 out tokens · 33250 ms · 2026-05-20T07:26:54.396879+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

109 extracted references · 109 canonical work pages · 17 internal anchors

  1. [1]

    2020 , booktitle =

    Chen, Andrew and Chow, Andy and Davidson, Aaron and DCunha, Arjun and Ghodsi, Ali and Hong, Sue Ann and Konwinski, Andy and Mewald, Clemens and Murching, Siddharth and Nykodym, Tomas and Ogilvie, Paul and Parkhe, Mani and Singh, Avesh and Xie, Fen and Zaharia, Matei and Zang, Richard and Zheng, Juntai and Zumar, Corey , month =. 2020 , booktitle =

    work page 2020

  2. [2]

    2024 , journal =

    Parker, Liam and Lanusse, Francois and Golkar, Siavash and Sarra, Leopoldo and Cranmer, Miles and Bietti, Alberto and Eickenberg, Michael and Krawezik, Geraud and McCabe, Michael and Morel, Rudy and Ohana, Ruben and Pettee, Mariel and R. 2024 , journal =. doi:10.1093/mnras/stae1450 , issn =

    work page doi:10.1093/mnras/stae1450 2024

  3. [3]

    International conference on machine learning , pages=

    Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures , author=. International conference on machine learning , pages=. 2013 , organization=

    work page 2013

  4. [4]

    Experiment Tracking with Weights and Biases , year =

    work page

  5. [5]

    and Meyers, Joshua E

    Crenshaw, John Franklin and Connolly, Andrew J. and Meyers, Joshua E. and Kalmbach, J. Bryce and Megias Homar, Guillem and Ribeiro, Tiago and Suberlak, Krzysztof and Thomas, Sandrine and Tsai, Te-Wei , number =. 2024 , journal =. doi:10.3847/1538-3881/ad1661 , issn =

    work page doi:10.3847/1538-3881/ad1661 2024

  6. [6]

    The redMaPPer Galaxy Cluster Catalog From DES Science Verification Data

    The RedMaPPer Galaxy Cluster Catalog From DES Science Verification Data. , keywords =. doi:10.3847/0067-0049/224/1/1 , archivePrefix =. 1601.00621 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0067-0049/224/1/1

  7. [7]

    and Gravet, R

    Huertas-Company, M. and Gravet, R. and Cabrera-Vives, G. and P. 2015 , journal =. doi:10.1088/0067-0049/221/1/8 , issn =

    work page doi:10.1088/0067-0049/221/1/8 2015

  8. [8]

    and Chen, A

    Zaharia, M. and Chen, A. and Davidson, A. and Ghodsi, A. and Hong, S. and Konwinski, A. and Murching, Siddharth and Nykodym, Tomas and Ogilvie, Paul and Parkhe, Mani and Xie, Fen and Zumar, Corey , pages =. 2018 , journal =

    work page 2018

  9. [9]

    Megan and Wang, Zhengdong and Schawinski, Kevin and Turp, Dennis and Powell, Meredith C

    Ghosh, Aritra and Urry, C. Megan and Wang, Zhengdong and Schawinski, Kevin and Turp, Dennis and Powell, Meredith C. , number =. 2020 , journal =. doi:10.3847/1538-4357/ab8a47 , issn =

    work page doi:10.3847/1538-4357/ab8a47 2020

  10. [10]

    , month =

    Hoyle, B. , month =. 2016 , journal =. doi:10.1016/J.ASCOM.2016.03.006 , issn =

    work page doi:10.1016/j.ascom.2016.03.006 2016

  11. [11]

    2015 , journal =

    Dieleman, Sander and Willett, Kyle W and Dambre, Joni , number =. 2015 , journal =. doi:10.1093/mnras/stv632 , keywords =

    work page doi:10.1093/mnras/stv632 2015

  12. [12]

    Ashley and Cranmer, Miles and Berger, Edo and Contardo, Gabriella and Ho, Shirley and Hosseinzadeh, Griffin and Lin, Joshua Yao-Yu , number =

    Villar, V. Ashley and Cranmer, Miles and Berger, Edo and Contardo, Gabriella and Ho, Shirley and Hosseinzadeh, Griffin and Lin, Joshua Yao-Yu , number =. 2021 , journal =. doi:10.3847/1538-4365/ac0893 , issn =

    work page doi:10.3847/1538-4365/ac0893 2021

  13. [13]

    , month =

    Fotopoulou, S. , month =. 2024 , journal =. doi:10.1016/j.ascom.2024.100851 , issn =

    work page doi:10.1016/j.ascom.2024.100851 2024

  14. [14]

    2021 , journal =

    Storey-Fisher, Kate and Huertas-Company, Marc and Ramachandra, Nesar and Lanusse, Francois and Leauthaud, Alexie and Luo, Yifei and Huang, Song and Prochaska, J Xavier , number =. 2021 , journal =. doi:10.1093/mnras/stab2589 , issn =

    work page doi:10.1093/mnras/stab2589 2021

  15. [15]

    Malanchev, K. L. and Pruzhinskaya, M. V. and Korolev, V. S. and Aleo, P. D. and Kornilov, M. V. and Ishida, E. E.O. and Krushinsky, V. V. and Mondon, F. and Sreejith, S. and Volnova, A. A. and Belinski, A. A. and Dodin, A. V. and Tatarnikov, A. M. and Zheltoukhov, S. G. , number =. 2021 , journal =. doi:10.1093/mnras/stab316 , issn =

    work page doi:10.1093/mnras/stab316 2021

  16. [16]

    and Bloom, J

    Rizhko, M. and Bloom, J. S. , number =. 2025 , journal =. doi:10.3847/1538-3881/adcbad , issn =

    work page doi:10.3847/1538-3881/adcbad 2025

  17. [17]

    2025 , journal =

    Lochner, Michelle and Rudnick, Lawrence , number =. 2025 , journal =. doi:10.3847/1538-3881/ada14c , issn =

    work page doi:10.3847/1538-3881/ada14c 2025

  18. [18]

    and Rotman, M

    Reis, I. and Rotman, M. and Poznanski, D. and Prochaska, J.X. and Wolf, L. , month =. 2021 , journal =. doi:10.1016/j.ascom.2020.100437 , issn =

    work page doi:10.1016/j.ascom.2020.100437 2021

  19. [19]

    2024 , journal =

    Mohale, Koketso and Lochner, Michelle , number =. 2024 , journal =. doi:10.1093/mnras/stae926 , issn =

    work page doi:10.1093/mnras/stae926 2024

  20. [20]

    2020 , journal =

    Walmsley, Mike and Smith, Lewis and Lintott, Chris and Gal, Yarin and Bamford, Steven and Dickinson, Hugh and Fortson, Lucy and Kruk, Sandor and Masters, Karen and Scarlata, Claudia and Simmons, Brooke and Smethurst, Rebecca and Wright, Darryl , number =. 2020 , journal =. doi:10.1093/mnras/stz2816 , issn =

    work page doi:10.1093/mnras/stz2816 2020

  21. [21]

    Ghosh, Aritra and Urry, C. Megan and Rau, Amrit and Perreault-Levasseur, Laurence and Cranmer, Miles and Schawinski, Kevin and Stark, Dominic and Tian, Chuan and Ofman, Ryan and Ananna, Tonima Tasnim and Auge, Connor and Cappelluti, Nico and Sanders, David B. and Treister, Ezequiel , number =. 2022 , journal =. doi:10.3847/1538-4357/ac7f9e , issn =

    work page doi:10.3847/1538-4357/ac7f9e 2022

  22. [22]

    and Vanderburg, Andrew , number =

    Shallue, Christopher J. and Vanderburg, Andrew , number =. 2018 , journal =. doi:10.3847/1538-3881/aa9e09 , issn =

    work page doi:10.3847/1538-3881/aa9e09 2018

  23. [23]

    2019 , journal =

    Ivezi. 2019 , journal =. doi:10.3847/1538-4357/ab042c , issn =

    work page doi:10.3847/1538-4357/ab042c 2019

  24. [24]

    2024 , journal =

    Zhang, Gemma and Helfer, Thomas and Gagliano, Alexander T and Mishra-Sharma, Siddharth and Ashley Villar, V , number =. 2024 , journal =. doi:10.1088/2632-2153/ad990d , issn =

    work page doi:10.1088/2632-2153/ad990d 2024

  25. [25]

    2019 , booktitle =

    Akiba, Takuya and Sano, Shotaro and Yanase, Toshihiko and Ohta, Takeru and Koyama, Masanori , month =. 2019 , booktitle =

    work page 2019

  26. [26]

    Ansel, Jason and Yang, Edward and He, Horace and Gimelshein, Natalia and Jain, Animesh and Voznesensky, Michael and Bao, Bin and Bell, Peter and Berard, David and Burovski, Evgeni and Chauhan, Geeta and Chourdia, Anjali and Constable, Will and Desmaison, Alban and DeVito, Zachary and Ellison, Elias and Feng, Will and Gong, Jiong and Gschwind, Michael and ...

    work page 2024

  27. [27]

    2022 , journal =

    Muthukrishna, Daniel and Mandel, Kaisey S and Lochner, Michelle and Webb, Sara and Narayan, Gautham , number =. 2022 , journal =. doi:10.1093/mnras/stac2582 , issn =

    work page doi:10.1093/mnras/stac2582 2022

  28. [28]

    and Lanusse, F

    Huertas-Company, M. and Lanusse, F. , month =. 2023 , journal =. doi:10.1017/pasa.2022.55 , issn =

    work page doi:10.1017/pasa.2022.55 2023

  29. [29]

    Megan and Ghosh, Aritra and Ofman, Ryan and Ananna, Tonima Tasnim and Auge, Connor and Cappelluti, Nico and Powell, Meredith C

    Tian, Chuan and Urry, C. Megan and Ghosh, Aritra and Ofman, Ryan and Ananna, Tonima Tasnim and Auge, Connor and Cappelluti, Nico and Powell, Meredith C. and Sanders, David B. and Schawinski, Kevin and Stark, Dominic and Tremblay, Grant R. , number =. 2023 , journal =. doi:10.3847/1538-4357/acad79 , issn =

    work page doi:10.3847/1538-4357/acad79 2023

  30. [30]

    and Scaife, Anna M

    Walmsley, Mike and Allen, Campbell and Aussel, Ben and Bowles, Micah and Gregorowicz, Kasia and Slijepcevic, Inigo Val and Lintott, Chris J. and Scaife, Anna M. M. and Jab. 2023 , journal =. doi:10.21105/joss.05312 , issn =

    work page doi:10.21105/joss.05312 2023

  31. [31]

    , keywords =

    Kuiper Belt Objects: Relics from the Accretion Disk of the Sun. , keywords =. doi:10.1146/annurev.astro.40.060401.093818 , adsurl =

    work page doi:10.1146/annurev.astro.40.060401.093818

  32. [32]

    and Gomes, R

    Tsiganis, K. and Gomes, R. and Morbidelli, A. and Levison, H. F. , date =. Origin of the orbital architecture of the giant planets of the Solar System , url =. Nature , number =. 2005 , bdsk-url-1 =. doi:10.1038/nature03539 , id =

    work page doi:10.1038/nature03539 2005

  33. [33]

    and Gerdes, David W

    Trilling, David E. and Gerdes, David W. and Juri. The Astronomical Journal , year =. doi:10.3847/1538-3881/ad1529 , archivePrefix =

    work page doi:10.3847/1538-3881/ad1529

  34. [34]

    and Kalmbach, J

    Whidden, Peter J. and Kalmbach, J. Bryce and Connolly, Andrew J. and Jones, R. Lynne and Smotherman, Hayden and Bektesevic, Dino and Slater, Colin and Becker, Andrew C. and Ivezić, Željko and Jurić, Mario and Bolin, Bryce and Moeyens, Joachim and Förster, Francisco and Golkhou, V. Zach , year=. Fast Algorithms for Slow Moving Asteroids: Constraints on the...

    work page doi:10.3847/1538-3881/aafd2d

  35. [35]

    and Kalmbach, J

    Smotherman, Hayden and Connolly, Andrew J. and Kalmbach, J. Bryce and Portillo, Stephen K. N. and Bektesevic, Dino and Eggl, Siegfried and Juric, Mario and Moeyens, Joachim and Whidden, Peter J. , year=. Sifting through the Static: Moving Object Detection in Difference Images , volume=. The Astronomical Journal , publisher=. doi:10.3847/1538-3881/ac22ff , number=

    work page doi:10.3847/1538-3881/ac22ff

  36. [36]

    2025 , eprint=

    Applying multimodal learning to Classify transient Detections Early (AppleCiDEr) I: Data set, methods, and infrastructure , author=. 2025 , eprint=

    work page 2025

  37. [37]

    2025 , eprint=

    BOOM and Babamul: a real-time, multi-survey, optical alert broker system operating at scale , author=. 2025 , eprint=

    work page 2025

  38. [38]

    2025 , eprint=

    AppleCiDEr II: SpectraNet -- A Deep Learning Network for Spectroscopic Data , author=. 2025 , eprint=

    work page 2025

  39. [39]

    and Drlica-Wagner, A

    Tanoglidis, D. and Drlica-Wagner, A. and Wei, K. and Li, T. S. and S. 2021 , journal =. doi:10.3847/1538-4365/abca89 , issn =

    work page doi:10.3847/1538-4365/abca89 2021

  40. [40]

    Hinton, G. E. and Salakhutdinov, R. R. , number =. 2006 , journal =. doi:10.1126/science.1127647 , issn =

    work page doi:10.1126/science.1127647 2006

  41. [41]

    2020 , booktitle =

    Grill, Jean-Bastien and Strub, Florian and Altch. 2020 , booktitle =

    work page 2020

  42. [42]

    2020 , booktitle =

    Chen, Ting and Kornblith, Simon and Norouzi, Mohammad and Hinton, Geoffrey , editor =. 2020 , booktitle =

    work page 2020

  43. [43]

    and Bechtol, K

    Sevilla-Noarbe, I. and Bechtol, K. and Carrasco Kind, M. and Carnero Rosell, A. and Becker, M. R. and Drlica-Wagner, A. and Gruendl, R. A. and Rykoff, E. S. and Sheldon, E. and Yanny, B. and Alarcon, A. and Allam, S. and Amon, A. and Benoit-L. 2021 , journal =. doi:10.3847/1538-4365/abeb66 , issn =

    work page doi:10.3847/1538-4365/abeb66 2021

  44. [44]

    doi:10.71929/RUBIN/2570308 , keywords =

    2025 , author =. doi:10.71929/RUBIN/2570308 , keywords =

    work page doi:10.71929/rubin/2570308 2025

  45. [45]

    and Hinton, Geoffrey E

    van der Maaten, Laurens J.P. and Hinton, Geoffrey E. , title =. Journal of Machine Learning Research , year =

    work page

  46. [46]

    Malkov, Yu. A. and Yashunin, D. A. , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

    work page

  47. [47]

    A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise , booktitle =

    Ester, Martin and Kriegel, Hans-Peter and Sander, J. A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise , booktitle =. 1996 , pages =

    work page 1996

  48. [48]

    Malanchev, Konstantin and DeLucchi, Melissa and Caplar, Neven and Malz, Alex I. and. Variability-finding in. arXiv e-prints , year =. 2506.23955 , archivePrefix =

    work page arXiv

  49. [49]

    arXiv e-prints , year =

    Caplar, Neven and Beebe, Wilson and Branton, Doug and Campos, Sandro and Connolly, Andrew and DeLucchi, Melissa and Jones, Derek and Juric, Mario and Kubica, Jeremy and Malanchev, Konstantin and Mandelbaum, Rachel and McGuire, Sean , title =. arXiv e-prints , year =. 2501.02103 , archivePrefix =

    work page arXiv

  50. [50]

    and others , title =

    Zhang, T. and others , title =. arXiv e-prints , year =. 2510.07370 , archivePrefix =

    work page arXiv

  51. [51]

    and Cristiani, S

    Arnouts, S. and Cristiani, S. and Moscardini, L. and Matarrese, S. and Lucchin, F. and Fontana, A. and Giallongo, E. , title =. Monthly Notices of the Royal Astronomical Society , year =

    work page

  52. [52]

    and Arnouts, S

    Ilbert, O. and Arnouts, S. and McCracken, H. J. and Bolzonella, M. and Bertin, E. and. Accurate photometric redshifts for the. Astronomy & Astrophysics , year =

    work page

  53. [53]

    R., Millman, K

    Harris, Charles R. and Millman, K. Jarrod and van der Walt, Stéfan J. and Gommers, Ralf and Virtanen, Pauli and Cournapeau, David and Wieser, Eric and Taylor, Julian and Berg, Sebastian and Smith, Nathaniel J. and Kern, Robert and Picus, Matti and Hoyer, Stephan and van Kerkwijk, Marten H. and Brett, Matthew and Haldane, Allan and del R. 2020 , journal =....

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

  54. [54]

    and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, Stéfan J

    Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, Stéfan J. and Brett, Matthew and Wilson, Joshua and Millman, K. Jarrod and Mayorov, Nikolay and Nelson, Andrew R. J. and Jones, Eric and K...

    work page doi:10.1038/s41592-019-0686-2 2020

  55. [55]

    Astropy: A Community Python Package for Astronomy

    Astropy: A community Python package for astronomy. , keywords =. 2013. doi:10.1051/0004-6361/201322068 , archivePrefix =. 1307.6212 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201322068 2013

  56. [56]

    The Astropy Project: Building an inclusive, open-science project and status of the v2.0 core package

    The Astropy Project: Building an Open-science Project and Status of the v2.0 Core Package. , keywords =. doi:10.3847/1538-3881/aabc4f , archivePrefix =. 1801.02634 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/aabc4f

  57. [57]

    D., 2007, @doi [Computing in Science & Engineering] 10.1109/MCSE.2007.55 , 9, 90

    Hunter, John D. , number =. 2007 , journal =. doi:10.1109/MCSE.2007.55 , issn =

    work page doi:10.1109/mcse.2007.55 2007

  58. [58]

    Research Notes of the AAS , volume=

    A Python Project Template for Healthy Scientific Software , author=. Research Notes of the AAS , volume=. 2024 , publisher=

    work page 2024

  59. [59]

    MNRAS , volume=

    Real-bogus classification for the Zwicky Transient Facility using deep learning , author=. MNRAS , volume=. 2019 , doi=

    work page 2019

  60. [60]

    van der Walt and Arien Crellin-Quick and Joshua S

    Stéfan J. van der Walt and Arien Crellin-Quick and Joshua S. Bloom , title =. 2019 , publisher =. doi:10.21105/joss.01247 , url =

    work page doi:10.21105/joss.01247 2019

  61. [61]

    W., Bloom, J

    A Data Science Platform to Enable Time-domain Astronomy. , keywords =. doi:10.3847/1538-4365/acdee1 , archivePrefix =. 2305.00108 , primaryClass =

    work page doi:10.3847/1538-4365/acdee1

  62. [62]

    , keywords =

    ViT-based Local Volume Dwarf Galaxy Identification (VIDA) in the CSST survey. , keywords =. doi:10.1093/mnras/staf1586 , archivePrefix =. 2506.04361 , primaryClass =

    work page doi:10.1093/mnras/staf1586

  63. [63]

    , keywords =

    Wide-field Survey of Dwarf Satellite Systems around 10 Hosts in the Local Volume. , keywords =. doi:10.3847/1538-4357/ab7758 , archivePrefix =. 1909.07389 , primaryClass =

    work page doi:10.3847/1538-4357/ab7758 1909

  64. [64]

    arXiv e-prints , keywords =

    SEAMLESS Survey: Four Faint Dwarf Galaxies Tracing Low-Mass Galaxy Evolution Across Environments. arXiv e-prints , keywords =. doi:10.48550/arXiv.2512.02118 , archivePrefix =. 2512.02118 , primaryClass =

    work page doi:10.48550/arxiv.2512.02118

  65. [65]

    , keywords =

    The Exploration of Local VolumE Satellites (ELVES) Survey: A Nearly Volume-limited Sample of Nearby Dwarf Satellite Systems. , keywords =. doi:10.3847/1538-4357/ac6fd7 , archivePrefix =. 2203.00014 , primaryClass =

    work page doi:10.3847/1538-4357/ac6fd7

  66. [66]

    arXiv e-prints , keywords =

    ELVES-Dwarf I: Satellites Systems of Eight Isolated Dwarf Galaxies in the Local Volume. arXiv e-prints , keywords =. doi:10.48550/arXiv.2504.08030 , archivePrefix =. 2504.08030 , primaryClass =

    work page doi:10.48550/arxiv.2504.08030

  67. [67]

    , keywords =

    Resolved Dwarf Galaxy Searches within 5 Mpc with the Vera Rubin Observatory and Subaru Hyper Suprime-Cam. , keywords =. doi:10.3847/1538-4357/ac0db8 , archivePrefix =. 2105.01658 , primaryClass =

    work page doi:10.3847/1538-4357/ac0db8

  68. [68]

    Milky Way Satellite Census. II. Galaxy-Halo Connection Constraints Including the Impact of the Large Magellanic Cloud. , keywords =. doi:10.3847/1538-4357/ab846a , archivePrefix =. 1912.03303 , primaryClass =

    work page doi:10.3847/1538-4357/ab846a 1912

  69. [69]

    , keywords =

    Quenching of satellite galaxies of Milky Way analogues: reconciling theory and observations. , keywords =. doi:10.1093/mnras/stac183 , archivePrefix =. 2109.06215 , primaryClass =

    work page doi:10.1093/mnras/stac183

  70. [70]

    Nature Astronomy , keywords =

    Baryonic solutions and challenges for cosmological models of dwarf galaxies. Nature Astronomy , keywords =. doi:10.1038/s41550-022-01689-w , archivePrefix =. 2206.05295 , primaryClass =

    work page doi:10.1038/s41550-022-01689-w

  71. [71]

    The Suppression of Star Formation on the Smallest Scales: What Role Does Environment Play?

    The suppression of star formation on the smallest scales: what role does environment play?. , keywords =. doi:10.1093/mnras/sty3357 , archivePrefix =. 1806.07891 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/sty3357

  72. [72]

    , keywords =

    EDGE: from quiescent to gas-rich to star-forming low-mass dwarf galaxies. , keywords =. doi:10.1093/mnras/staa1640 , archivePrefix =. 2004.09530 , primaryClass =

    work page doi:10.1093/mnras/staa1640 2004

  73. [73]

    , keywords =

    Small-Scale Challenges to the CDM Paradigm. , keywords =. doi:10.1146/annurev-astro-091916-055313 , archivePrefix =. 1707.04256 , primaryClass =

    work page doi:10.1146/annurev-astro-091916-055313

  74. [74]

    , keywords =

    The DECam Local Volume Exploration Survey: Overview and First Data Release. , keywords =. doi:10.3847/1538-4365/ac079d , archivePrefix =. 2103.07476 , primaryClass =

    work page doi:10.3847/1538-4365/ac079d

  75. [75]

    , keywords =

    Third data release of the Hyper Suprime-Cam Subaru Strategic Program. , keywords =. doi:10.1093/pasj/psab122 , archivePrefix =. 2108.13045 , primaryClass =

    work page doi:10.1093/pasj/psab122

  76. [76]

    Rubin Observatory Science Pipelines Developers

    Vera C. Rubin Observatory Science Pipelines Developers. The LSST Science Pipelines Software: Optical Survey Pipeline Reduction and Analysis Environment. 2025. doi:10.71929/rubin/2570545

    work page doi:10.71929/rubin/2570545 2025

  77. [77]

    Deep Residual Learning for Image Recognition

    Deep Residual Learning for Image Recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year = 2016, month = jun, eid =. doi:10.1109/CVPR.2016.90 , archivePrefix =. 1512.03385 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1109/cvpr.2016.90 2016

  78. [78]

    Line-of-Sight Structure Toward Strong Lensing Galaxy Clusters

    Line-of-sight Structure toward Strong Lensing Galaxy Clusters. , keywords =. doi:10.1088/0004-637X/783/1/41 , archivePrefix =. 1312.3637 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-637x/783/1/41

  79. [79]

    COOL-LAMPS. I. An Extraordinarily Bright Lensed Galaxy at Redshift 5.04. , keywords =. doi:10.3847/1538-4357/abcb86 , archivePrefix =. 2011.06601 , primaryClass =

    work page doi:10.3847/1538-4357/abcb86 2011

  80. [80]

    COOL-LAMPS. II. Characterizing the Size and Star Formation History of a Bright Strongly Lensed Early-type Galaxy at Redshift 1.02. , keywords =. doi:10.3847/1538-4357/ac9974 , archivePrefix =. 2203.11957 , primaryClass =

    work page doi:10.3847/1538-4357/ac9974

Showing first 80 references.