Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

A bidirectional Brownian-bridge diffusion model learns the probabilistic mapping between LSST and Euclid images, enabling both missing-survey synthesis and rare-event detection from reconstruction inconsistency.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-14 22:31 UTC pith:NRR2SSFH

load-bearing objection Solid sim-only feasibility paper for probabilistic LSST–Euclid translation; anomaly gains are real but OOD-by-construction, so treat pipeline claims as provisional. the 3 major comments →

arxiv 2603.11928 v2 pith:NRR2SSFH submitted 2026-03-12 astro-ph.IM cs.CV

AS-Bridge: A Bidirectional Generative Framework Bridging Next-Generation Astronomical Surveys

Dichang Zhang , Yixuan Shao , Simon Birrer , Dimitris Samaras This is my paper
classification astro-ph.IM cs.CV
keywords observational cosmologyastronomical surveygenerative modelmultimodal learningBrownian bridgeLSSTEuclidanomaly detection
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

LSST and Euclid will map overlapping sky with different resolutions, bands, noise, and cadences, so the same galaxy looks systematically different in each survey and a single deterministic map between them is ill-posed. AS-Bridge treats the two observations as endpoints of a stochastic Brownian bridge and trains a diffusion model on paired cutouts so that either survey can be sampled from the other while preserving calibrated uncertainty. On forward-modeled data the method recovers the target conditional distribution better than standard image-to-image baselines and, when both surveys are available, flags rare systems such as strong gravitational lenses by the failure of multi-sample reconstructions to match the true image. The authors argue this inter-survey generative layer can fill gaps in non-overlapping sky regions and act as an unsupervised discovery tool once real joint data arrive.

Core claim

AS-Bridge shows that a bidirectional Brownian-bridge diffusion model, trained with an ε-prediction objective on paired LSST–Euclid cutouts, recovers the conditional distribution of one survey given the other more faithfully than competing generative translators and converts multi-sample reconstruction inconsistency into a practical anomaly score for rare events such as strong lenses.

What carries the argument

Bidirectional Brownian bridge diffusion: a stochastic process whose endpoints are the two survey observations; the reverse dynamics are trained so that, from any intermediate noisy fusion, either endpoint can be reconstructed while the learned score matches the true conditional.

Load-bearing premise

That the forward-modeled simulated cutouts already capture the real conditional relationship between LSST and Euclid well enough for the reported translation and detection gains to transfer once official survey data exist.

What would settle it

Train and evaluate the same bridge on real overlapping LSST–Euclid cutouts after official releases; if CRPS relative to baselines and strong-lens AUPR collapse, the sim-to-real claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces AS-Bridge, a bidirectional generative model that treats translation between ground-based LSST (g,r,i) and space-based Euclid (VIS) imaging as a stochastic Brownian-bridge diffusion process anchored on overlapping-sky paired cutouts. After training on forward-modeled SLSim regular-galaxy pairs, the model is used for (i) probabilistic inter-survey restoration/synthesis evaluated by CRPS and (ii) unsupervised rare-event detection via multi-sample min reconstruction error on the Euclid side after bridge fusion, with held-out galaxy–galaxy strong lenses as the anomaly class. On the simulated benchmark, ε-prediction AS-Bridge reports the best CRPS in both directions versus SPADE, OASIS, pix2pix, Palette, and joint diffusion (Table 1), and stronger precision-oriented detection metrics than Deco-Diff and CFM (Table 2: AUPR 0.80, FPR@1% TPR = 0%). The authors frame the work as a simulation-first proof of concept for future LSST–Euclid joint pipelines, with code and data released.

Significance. If the reported conditional modeling and cross-survey inconsistency scores transfer beyond the present simulations, AS-Bridge would be a concrete, reusable component for joint LSST–Euclid analysis: filling non-overlapping footprints under uncertainty, deblending under atmospheric seeing, and flagging structural outliers without survey-specific anomaly labels. Strengths that should be credited include a clear problem formulation of bidirectional ill-posed survey translation, a likelihood-motivated ε-prediction objective with a short formal argument (Proposition 1), standardized CRPS and discovery-oriented AD metrics, systematic baselines, and public code/data. The work is timely for the Rubin–Euclid era and establishes a useful simulated benchmark even if real-data validation remains future work.

major comments (3)
  1. [§3.3, Table 2] §3.3, Eqs. (19)–(20), §4.1, and Table 2: The rare-event detector is trained only on 110k regular galaxies; all 5k strong lenses are held out and generated by a distinct SLSim process (deflector–source pairing, Einstein-ring morphology) never present in training. Elevated multi-sample reconstruction error is therefore expected by construction as an OOD morphology test, not as evidence that the same score would flag true rare events that still arise from the regular-galaxy generative family (unusual SEDs, mergers, artifacts, or rare but in-family morphologies). Because inter-survey rare-event detection is presented as a load-bearing scientific capability (§1, §6), the manuscript should either (a) reframe Table 2 as OOD structural-outlier detection and temper pipeline claims, or (b) add experiments that inject rare but in-distribution deviations (e.g., atypical Sérsic parameters, color outl
  2. [§4.1, §5] §4.1 and §5: All quantitative claims rest on SLSim forward models with survey-specific PSFs, bandpasses, and noise. The abstract and conclusion position AS-Bridge as complementary to future LSST–Euclid joint data pipelines once real releases exist. That positioning is only weakly supported: no domain-randomization, no controlled sim-to-real stress tests (PSF mismatch, background, calibration residuals, bandpass truncation effects already noted for VIS), and no partial real-data pilot (e.g., Euclid Q1 / early Rubin preview cutouts where available). The central feasibility claim on the simulated benchmark can stand, but the pipeline-readiness language should be narrowed to what the experiments actually show, or additional transfer diagnostics should be added.
  3. [§3.2.2] §3.2.2: The VE-style likelihood weighting and Proposition 1 rely on the assumption that marginal data variance is identical across LSST and Euclid domains. Real LSST multi-band optical and Euclid VIS cutouts differ in dynamic range, noise structure, and pixel scale after resampling; the paper does not report empirical variance matching, per-channel normalization, or sensitivity of CRPS/AD scores when this assumption is violated. A short ablation (rescale one domain’s variance; recompute Table 1) would show whether the ε-prediction advantage is robust or tied to the simulation’s matched rendering.
minor comments (6)
  1. [§3.1] §3.1: The two conditionals are written twice as p(x_Euclid | x_LSST); the second should be p(x_LSST | x_Euclid).
  2. [Table 1] Table 1 header cites SPADE[20]; the SPADE reference is Park et al. [27] in the bibliography. Align citation numbers.
  3. [Figures 1–4] Fig. 1 caption in the body refers to midpoint reconstructions, but the overview figure description earlier mixes footprint and method schematic; ensure figure numbering and captions match the qualitative panels (Figs. 2–4).
  4. [§4.3.1] §4.3.1: “Contunuous Ranked Probability Score” → Continuous. Also state the number of CRPS samples M and AD samples N used for Tables 1–2 for reproducibility.
  5. [Eq. (4)] Eq. (4) writes x_t = x_0 + m_t (x_t − x_0) + …; the middle term should involve (x_T − x_0), consistent with Eq. (5).
  6. [§4.1.5] Clarify cutout resampling: 64×64 for both surveys with different native pixel scales implies different physical FOV; state the common angular FOV used for pairing.

Circularity Check

0 steps flagged

No circular derivation: empirical generative modeling and standard unsupervised AD evaluation on held-out simulated data with external proper scores.

full rationale

AS-Bridge is an empirical ML methods paper. The Brownian-bridge formulation, reverse process, and ε-prediction objective (Prop. 1 equating it to a milder √δ_t-weighted score-matching loss) are derived from standard diffusion/bridge literature (Ho et al., Song et al., BBDM) and then trained/evaluated; nothing is defined in terms of the reported CRPS or AUPR. Translation quality is measured by CRPS (a strictly proper external scoring rule) against held-out regular-galaxy cutouts and against independent baselines (SPADE, OASIS, pix2pix, Palette, joint diffusion). Rare-event detection trains exclusively on regular galaxies and scores held-out strong lenses via multi-sample min reconstruction error (Eqs. 19–20); this is ordinary unsupervised anomaly detection, not a self-definitional or fitted-input “prediction.” Lenses are OOD by construction of the SLSim pipeline, which is a validity/sim-to-real concern, not circularity of the derivation chain. No load-bearing uniqueness theorem or ansatz is imported from overlapping-author prior work. The paper is self-contained against its own simulated benchmark and external metrics; score 0 is the correct non-finding.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 1 invented entities

The central claims rest on standard diffusion/bridge math, survey observation operators, and simulation fidelity rather than new physical entities. Load-bearing modeling choices are the identical-marginal-variance (VE) assumption for the two surveys, the unsupervised reconstruction-failure model of rarity, and the fidelity of SLSim forward modeling. Free parameters are ordinary ML/AD knobs (timesteps, sample count N, network capacity) not cosmology constants fitted to the target metrics.

free parameters (4)
  • Brownian bridge / diffusion timestep schedule and noise scale δ_t
    Controls the forward fusion and reverse sampling path; standard diffusion hyperparameter not derived from first principles for LSST–Euclid.
  • Number of stochastic reconstructions N for anomaly map
    Anomaly score takes min error over N samples (§3.3); N is a design choice that affects noise suppression vs compute.
  • Neural network architecture and training hyperparameters for f_θ / ε_θ
    Capacity and optimization settings determine empirical CRPS/AUPR; not uniquely fixed by the problem statement.
  • Simulation selection cuts (magnitude, redshift, image separation, detectability)
    Catalog and lens filters in §4.1 shape the training and evaluation distributions and thus reported metrics.
axioms (5)
  • domain assumption Each survey observation is a noisy, band-limited, PSF-convolved realization of an unobservable latent scene Φ; cross-survey maps are stochastic conditionals, not deterministic functions.
    §3.1 equations (1)–(2); justifies generative rather than deterministic translation.
  • ad hoc to paper Marginal data variance is identical across LSST and Euclid domains, matching the variance-exploding diffusion assumption.
    Stated in §3.2.2 to import Song et al. likelihood weighting into the bridge; not independently verified on real survey pixels.
  • domain assumption Rare or unseen phenomena are underrepresented in training and therefore cannot be faithfully reconstructed, so reconstruction inconsistency is a valid anomaly score.
    §3.3 core assumption for unsupervised rare-event detection.
  • standard math Brownian bridge forward/reverse transitions and the ε-prediction equivalence (Proposition 1) hold under the stated Gaussian bridge process.
    §3.2.1–3.2.2; standard SDE/bridge identities plus a short algebraic proof.
  • domain assumption SLSim/SkyPy forward models with chosen cosmology (H0=70, Ωm=0.3) and instrument transfer functions adequately stand in for real LSST/Euclid pairs for method development.
    §4.1 and §5; simulation-first practice acknowledged as expected domain gap.
invented entities (1)
  • AS-Bridge (bidirectional inter-survey Brownian bridge generative framework) no independent evidence
    purpose: Unified model for probabilistic LSST↔Euclid translation and reconstruction-based rare-event scoring.
    Methodological construct, not a new physical object; independent evidence would be real-survey performance, not yet shown.

pith-pipeline@v1.1.0-grok45 · 24043 in / 3390 out tokens · 37717 ms · 2026-07-14T22:31:14.522932+00:00 · methodology

0 comments
read the original abstract

The upcoming decade of observational cosmology will be shaped by large sky surveys, such as the ground-based LSST at the Vera C. Rubin Observatory and the space-based Euclid mission. While they promise an unprecedented view of the Universe across depth, resolution, and wavelength, their differences in observational modality, sky coverage, point-spread function, and scanning cadence make joint analysis beneficial, but also challenging. To facilitate joint analysis, we introduce A(stronomical)S(urvey)-Bridge, a bidirectional generative model that translates between ground- and space-based observations. AS-Bridge learns a diffusion model that employs a stochastic Brownian Bridge process between the LSST and Euclid observations. The two surveys have overlapping sky regions, where we can explicitly model the conditional probabilistic distribution between them. We show that this formulation enables new scientific capabilities beyond single-survey analysis, including faithful probabilistic predictions of missing survey observations and inter-survey detection of rare events. These results establish the feasibility of inter-survey generative modeling. AS-Bridge is therefore well-positioned to serve as a complementary component of future LSST-Euclid joint data pipelines, enhancing the scientific return once data from both surveys become available. Data and code are available at https://github.com/ZHANG7DC/AS-Bridge.

Figures

Figures reproduced from arXiv: 2603.11928 by Dichang Zhang, Dimitris Samaras, Simon Birrer, Yixuan Shao.

Figure 1
Figure 1. Figure 1: Overview of AS-Bridge. The central panel shows [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 4
Figure 4. Figure 4: Midpoint-to-Euclid AS-Bridge reconstructions for [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Euclid-to-LSST translation with multiple realiza [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. FluxFlow: Conservative Flow-Matching for Astronomical Image Super-Resolution

    cs.CV 2026-05 unverdicted novelty 7.0

    FluxFlow is a conservative pixel-space flow-matching framework for astronomical super-resolution that incorporates real atmospheric uncertainty and a training-free Wiener correction, outperforming baselines on a new 1...

  2. FluxFlow: Conservative Flow-Matching for Astronomical Image Super-Resolution

    cs.CV 2026-05 unverdicted novelty 5.0

    FluxFlow uses conservative pixel-space flow-matching with uncertainty weights and Wiener test-time correction to outperform baselines on photometric and scientific accuracy for ground-to-space super-resolution, valida...

Reference graph

Works this paper leans on

38 extracted references · 5 canonical work pages · cited by 1 Pith paper

  1. [1]

    Peter Ade, James Aguirre, Zeeshan Ahmed, Simone Aiola, Aamir Ali, David Alonso, Marcelo A. Alvarez, Kam Arnold, Peter Ashton, Jason Austermann, Humna Awan, Carlo Baccigalupi, Taylor Baildon, Darcy Barron, Nick Battaglia, Richard Battye, Eric Baxter, Andrew Bazarko, James A. Beall, Rachel Bean, Do- minic Beck, Shawn Beckman, Benjamin Beringue, Federico Bia...

    2019

  2. [2]

    de la Bella, Simon Birrer, Sarah Bridle, Juan Pablo Cordero, Ginevra Favole, Ian Harrison, Ian W

    Adam Amara, Lucia F. de la Bella, Simon Birrer, Sarah Bridle, Juan Pablo Cordero, Ginevra Favole, Ian Harrison, Ian W. Harry, William G. Hartley, Cole- man Krawczyk, Andrew Lundgren, Brian Nord, Laura K. Nuttall, Richard P. Rollins, Philipp Sudek, Sut-Ieng Tam, Nicolas Tessore, Arthur E. Tolley, Keiichi Umetsu, Andrew R. Williamson, and Laura Wolz. 2021. ...

    work page doi:10.21105/joss.03056 2021

  3. [3]

    Farzad Beizaee, Gregory A Lodygensky, Christian Desrosiers, and Jose Dolz

  4. [4]

    InProceedings of the Computer Vision and Pattern Recognition Conference

    Correcting deviations from normality: A reformulated diffusion model for multi-class unsupervised anomaly detection. InProceedings of the Computer Vision and Pattern Recognition Conference. 19088–19097

  5. [5]

    Blanchard, S

    A. Blanchard, S. Camera, C. Carbone, V. F. Cardone, S. Casas, S. Clesse, S. Ilić, M. Kilbinger, T. Kitching, M. Kunz, F. Lacasa, E. Linder, E. Majerotto, K. Markovič, M. Martinelli, V. Pettorino, A. Pourtsidou, Z. Sakr, A. G. Sánchez, D. Sapone, I. Tutusaus, S. Yahia-Cherif, V. Yankelevich, S. Andreon, H. Aussel, A. Balaguera- Antolínez, M. Baldi, S. Bard...

    doi:10.1051/0004-6361/202038071 2020

  6. [6]

    Goldstein, Isobel Hook, Željko Ivezić, Steven M

    The LSST Dark Energy Science Collaboration, Rachel Mandelbaum, Tim Ei- fler, Renée Hložek, Thomas Collett, Eric Gawiser, Daniel Scolnic, David Alonso, Humna Awan, Rahul Biswas, Jonathan Blazek, Patricia Burchat, Nora Elisa Chis- ari, Ian Dell’Antonio, Seth Digel, Josh Frieman, Daniel A. Goldstein, Isobel Hook, Željko Ivezić, Steven M. Kahn, Sowmya Kamath,...

    Pith/arXiv arXiv 2021

  7. [7]

    Alex Costanzino, Pierluigi Zama Ramirez, Giuseppe Lisanti, and Luigi Di Stefano

  8. [8]

    InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition. CVPR

  9. [9]

    Miguel Crispim Romão, Djuna Croon, and Daniel Godines. 2025. Anomaly detection to identify transients in LSST time series data.Monthly Notices of the Royal Astronomical Society543, 1 (09 2025), 351–357. arXiv:https://academic.oup.com/mnras/article- pdf/543/1/351/64222460/staf1477.pdf doi:10.1093/mnras/staf1477

    work page doi:10.1093/mnras/staf1477 2025

  10. [10]

    De Bonis, S

    N. De Bonis, S. Vaccaro, Y. Maruccia, G. Riccio, R. Crupi, S. Rubini, D. De Cicco, M. Brescia, and S. Cavuoti. 2026. Hunting the outliers: Machine learning for anomalous time series detection.Astronomy and Computing55 (2026), 101049. doi:10.1016/j.ascom.2025.101049

    work page doi:10.1016/j.ascom.2025.101049 2026

  11. [11]

    and DESC & SLSC Collaboration

    Khadka et al. and DESC & SLSC Collaboration. 2025. SLSim: Strong Lensing Simulation pipeline. GitHub repository. https://github.com/LSST-strong-lensing/ slsim Documentation: https://slsim.readthedocs.io. Accessed: 2026-01-29

    2025

  12. [12]

    Mellier, Abdurro’uf, J

    Euclid Collaboration, Y. Mellier, Abdurro’uf, J. A. Acevedo Barroso, A. Achúcarro, J. Adamek, R. Adam, G. E. Addison, N. Aghanim, M. Aguena, V. Ajani, Y. Akrami, A. Al-Bahlawan, A. Alavi, I. S. Albuquerque, G. Alestas, G. Alguero, A. Allaoui, S. W. Allen, V. Allevato, A. V. Alonso-Tetilla, B. Altieri, A. Alvarez-Candal, S. Alvi, A. Amara, L. Amendola, J. ...

    2025

  13. [13]

    Yaoxuan Feng, Wenchao Chen, Yuxin Li, Bo Chen, Yubiao Wang, Zixuan Zhao, Hongwei Liu, and Mingyuan Zhou. 2025. OmiAD: One-Step Adaptive Masked Jan 2026, , Zhang et al. Diffusion Model for Multi-class Anomaly Detection via Adversarial Distillation. In Proceedings of the 42nd International Conference on Machine Learning (Proceedings of Machine Learning Rese...

    2025

  14. [14]

    Tilmann Gneiting and Adrian E Raftery. 2007. Strictly Proper Scoring Rules, Prediction, and Estimation.J. Amer. Statist. Assoc.102, 477 (2007), 359–378

    2007

  15. [15]

    Vavrek, Guillermo Buenadicha, John Hoar, Sandor Kruk, and Jan Reerink

    Pablo Gómez, Roland D. Vavrek, Guillermo Buenadicha, John Hoar, Sandor Kruk, and Jan Reerink. 2024. Machine learning-driven Anomaly Detection and Forecasting for Euclid Space Telescope Operations. arXiv:2411.05596 [cs.LG] https://arxiv.org/abs/2411.05596

    Pith/arXiv arXiv 2024

  16. [17]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models.arXiv preprint arxiv:2006.11239(2020)

    Pith/arXiv arXiv 2020

  17. [18]

    Jiahe Huang, Guandao Yang, Zichen Wang, and Jeong Joon Park. 2024. DiffusionPDE: Generative PDE-Solving Under Partial Observation. arXiv:2406.17763 [cs.LG]

    Pith/arXiv arXiv 2024

  18. [19]

    E. E. O. Ishida, M. V. Kornilov, K. L. Malanchev, M. V. Pruzhinskaya, A. A. Volnova, V. S. Korolev, F. Mondon, S. Sreejith, A. A. Malancheva, and S. Das. 2021. Active anomaly detection for time-domain discoveries.Astronomy & Astrophysics650 (June 2021), A195. doi:10.1051/0004-6361/202037709

    work page doi:10.1051/0004-6361/202037709 2021

  19. [20]

    Phillip Isola, Jun-Yan Zhu, Tinghui Zhou, and Alexei A Efros. 2017. Image-to- Image Translation with Conditional Adversarial Networks.CVPR(2017)

    2017

  20. [21]

    Željko Ivezić, Steven M. Kahn, J. Anthony Tyson, Bob Abel, Emily Acosta, Robyn Allsman, David Alonso, Yusra AlSayyad, Scott F. Anderson, John Andrew, James Roger P. Angel, George Z. Angeli, Reza Ansari, Pierre Antilogus, Constanza Araujo, Robert Armstrong, Kirk T. Arndt, Pierre Astier, Éric Aubourg, Nicole Auza, Tim S. Axelrod, Deborah J. Bard, Jeff D. Ba...

    2019

  21. [22]

    2024.desihub/speclite: Bug fix release: General clean-up prior to refactoring package infrastructure (v0.20)

    David Kirkby et al. 2024.desihub/speclite: Bug fix release: General clean-up prior to refactoring package infrastructure (v0.20). doi:10.5281/zenodo.13225530

    doi:10.5281/zenodo.13225530 2024

  22. [23]

    Simon A. A. Kohl, Bernardino Romera-Paredes, Clemens Meyer, Jeffrey De Fauw, Joseph R. Ledsam, Klaus H. Maier-Hein, S. M. Ali Eslami, Danilo Jimenez Rezende, and Olaf Ronneberger. 2018. A Probabilistic U-Net for Segmentation of Ambigu- ous Images. InAdvances in Neural Information Processing Systems 31 (NeurIPS 2018)

    2018

  23. [24]

    Kurlander, Pedro H

    Jacob A. Kurlander, Pedro H. Bernardinelli, Megan E. Schwamb, Mario Jurić, Joseph Murtagh, Colin Orion Chandler, Stephanie R. Merritt, David Nesvorný, David Vokrouhlický, R. Lynne Jones, Grigori Fedorets, Samuel Cornwall, Matthew J. Holman, Siegfried Eggl, Drew Oldag, Maxine West, Jeremy Kubica, Peter Yoachim, Joachim Moeyens, Kathleen Kiker, and Laura E....

    doi:10.3847/1538-3881/add685 2025

  24. [25]

    Bo Li, Kaitao Xue, Bin Liu, and Yu-Kun Lai. 2023. BBDM: Image-to-image trans- lation with Brownian bridge diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1952–1961

    2023

  25. [26]

    Fei Tony Liu, Kai Ming Ting, and Zhi-Hua Zhou. 2008. Isolation Forest. In2008 Eighth IEEE International Conference on Data Mining. 413–422. doi:10.1109/ICDM. 2008.17

    doi:10.1109/icdm 2008

  26. [27]

    Zonglin Lyu, Ming Li, Jianbo Jiao, and Chen Chen. 2024. Frame Interpolation with Consecutive Brownian Bridge Diffusion. InProceedings of the 32nd ACM International Conference on Multimedia (MM ’24)

    2024

  27. [28]

    Rauscher, Chris Bennett, Edward S

    Gregory Mosby, Bernard J. Rauscher, Chris Bennett, Edward S. Cheng, Stephanie Cheung, Analia Cillis, David Content, Dave Cottingham, Roger Foltz, John Gygax, Robert J. Hill, Jeffrey W. Kruk, Jon Mah, Lane Meier, Chris Merchant, Laddawan Miko, Eric C. Piquette, Augustyn Waczynski, and Yiting Wen. 2020. Properties and characteristics of the Nancy Grace Roma...

    2020

  28. [29]

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. 2019. Semantic Image Synthesis with Spatially-Adaptive Normalization. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    2019

  29. [30]

    Stephan Rasp, Stephan Hoyer, Alexander Merose, Ian Langmore, Peter Battaglia, Tyler Russel, Alvaro Sanchez-Gonzalez, Vivian Yang, Rob Carver, Shreya Agrawal, Matthew Chantry, Zied Ben Bouallegue, Peter Dueben, Carla Bromberg, Jared Sisk, Luke Barrington, Aaron Bell, and Fei Sha. 2023. WeatherBench 2: A benchmark for the next generation of data-driven glob...

    Pith/arXiv arXiv 2023

  30. [31]

    Chitwan Saharia, William Chan, Huiwen Chang, Chris Lee, Jonathan Ho, Tim Salimans, David Fleet, and Mohammad Norouzi. 2022. Palette: Image-to-image diffusion models. InACM SIGGRAPH 2022 conference proceedings. 1–10

    2022

  31. [32]

    Edgar Schonfeld, Vadim Sushko, Dan Zhang, Juergen Gall, Bernt Schiele, and Anna Khoreva. 2021. You Only Need Adversarial Supervision for Semantic Image Synthesis. InInternational Conference on Learning Representations. https: //openreview.net/forum?id=yvQKLaqNE6M

    2021

  32. [33]

    Schwamb, R

    Megan E. Schwamb, R. Lynne Jones, Steven R. Chesley, Alan Fitzsimmons, Wes- ley C. Fraser, Matthew J. Holman, Henry Hsieh, Darin Ragozzine, Cristina A. Thomas, David E. Trilling, Michael E. Brown, Michele T. Bannister, Dennis Bode- wits, Miguel de Val-Borro, David Gerdes, Mikael Granvik, Michael S. P. Kelley, Matthew M. Knight, Robert L. Seaman, Quan-Zhi ...

    Pith/arXiv arXiv 2018

  33. [34]

    Yang Song, Conor Durkan, Iain Murray, and Stefano Ermon. 2021. Maximum Likelihood Training of Score-Based Diffusion Models. InThirty-Fifth Conference on Neural Information Processing Systems

    2021

  34. [35]

    Lynne Jones, and Peter Birtwhistle

    Tom Wagg, Mario Juric, Peter Yoachim, Jake Kurlander, Sam Cornwall, Joachim Moeyens, Siegfried Eggl, R. Lynne Jones, and Peter Birtwhistle. 2025. Expected Impact of Rubin Observatory LSST on NEO Follow-up.The Astronomical Journal 169, 6 (may 2025), 315. doi:10.3847/1538-3881/adc89f

    work page doi:10.3847/1538-3881/adc89f 2025

  35. [36]

    Yue Wang, Jinlong Peng, Jiangning Zhang, Ran Yi, Yabiao Wang, and Chengjie Wang. 2023. Multimodal Industrial Anomaly Detection via Hybrid Fusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 8032–8041

    2023

  36. [37]

    Julia Wolleb, Florentin Bieder, Robin Sandkühler, and Philippe C. Cattin. 2022. Dif- fusion Models for Medical Anomaly Detection. InMedical Image Computing and Computer Assisted Intervention – MICCAI 2022, Linwei Wang, Qi Dou, P. Thomas Fletcher, Stefanie Speidel, and Shuo Li (Eds.). Springer Nature Switzerland, Cham, 35–45

    2022

  37. [38]

    Hantao Zhang, Yuhe Liu, Jiancheng Yang, Weidong Guo, Xinyuan Wang, and Pascal Fua. 2025. DiffAtlas: GenAI-fying Atlas Segmentation via Image-Mask Diffusion . Inproceedings of Medical Image Computing and Computer Assisted Intervention – MICCAI 2025, Vol. LNCS 15975. Springer Nature Switzerland

    2025

  38. [39]

    2003.Stochastic Differential Equations: An Introduction with Applications

    Bernt Øksendal. 2003.Stochastic Differential Equations: An Introduction with Applications. Springer

    2003