Modelos de difusión probabilísticos como emuladores en simulaciones de gravedad modificada

dc.contributor.advisorHortua Orjuela, ‪Hector Javie
dc.contributor.authorSaavedra Tafur, Paola Andrea
dc.contributor.authorRiveros Galeano, Julieth Katherine
dc.date.accessioned2024-09-06T19:10:10Z
dc.date.available2024-09-06T19:10:10Z
dc.date.issued2024-06
dc.description.abstractEn este trabajo, se usa la aplicación de modelos de difusión probabilística (DDPM) y de eliminación de ruido (DDIM) en la generación de simulaciones de N-cuerpos en modelos de gravedad modificada. Los DDPM añaden ruido de manera controlada a través de una cadena de Markov, donde cada paso de difusión depende exclusivamente del anterior, incrementando gradualmente la indistinción. Por otro lado, los DDIM introducen una mayor flexibilidad al permitir referencias a estados anteriores más distantes en el proceso de difusión. En este documento reportamos que los modelos DDPM proveen un excelente emulador para la generación de las simulaciones cosmológicas a nivel del espectro de potencias y biespectro obteniendo un r2 = 0.8, mientras que los modelos DDIM fallan en la extracción de la normalización. Este enfoque juega un papel importante en cosmología, donde la precisión y la rapidez en la generación de simulaciones son esenciales para la estimación de parámetros y la restricción de modelos de gravedad.
dc.description.abstractenglishIn this work, the application of Denoising Diffusion Probabilistic Models (DDPM) and Denoising Diffusion Implicit Models (DDIM) is used in the generation of N-body simulations in modified gravity models. DDPMs add noise in a controlled manner through a Markov chain, where each diffusion step depends exclusively on the previous one, gradually increasing indistinguishability. On the other hand, DDIMs introduce greater flexibility by allowing references to more distant previous states in the diffusion process. In this document, we report that DDPM models provide an excellent emulator for the generation of cosmological simulations at the level of power spectrum and bispectrum, achieving an r2 = 0.8, while DDIM models fail in the extraction of normalization. This approach plays an important role in cosmology, where precision and speed in the generation of simulations are essential for parameter estimation and model constraint of gravity.
dc.identifier.urihttps://hdl.handle.net/20.500.12495/12962
dc.language.isoes
dc.relation.referencesAbadi M., contributors 2023, TensorFlow: An End-to-End Open Source Machine Learning Platform, https://github.com/ tensorflow/tensorflow
dc.relation.referencesAbareshi B., et al., 2022, The Astronomical Journal, 164, 207
dc.relation.referencesBernardeau F., Colombi S., Gaztanaga E., Scoccimarro R., 2002, Physics Reports, 367, 1
dc.relation.referencesB´eres A., 2022, Denoising Diffusion Implicit Models, https:// keras.io/examples/generative/ddim/
dc.relation.referencesChollet F., contributors 2023, Keras: The Python Deep Learning library, https://github.com/keras-team/keras
dc.relation.referencesCorso G., St¨ark H., Jing B., Barzilay R., Jaakkola T., 2023, Diff Dock: Diffusion Steps, Twists, and Turns for Molecular Dock ing (arXiv:2210.01776)
dc.relation.referencesGarc´ıa-Farieta J. E., Hort´ ua H. J., Kitaura F.-S., 2024, Astronomy & Astrophysics, 684, A100
dc.relation.referencesGoodfellow I., Pouget-Abadie J., Mirza M., Xu B., Warde-Farley D., Ozair S., Courville A., Bengio Y., 2014, Advances in neural information processing systems, 27
dc.relation.referencesHo J., Jain A., Abbeel P., 2020, arXiv preprint arxiv:2006.11239
dc.relation.referencesHu W., Sawicki I., 2007, Physical Review D, 76, 064004
dc.relation.referencesKingma D. P., Welling M., 2022, Auto-Encoding Variational Bayes (arXiv:1312.6114)
dc.relation.referencesKoda J., Blake C., Beutler F., Kazin E., Marin F., 2016, Monthly Notices of the Royal Astronomical Society, 459, 2118–2129
dc.relation.referencesKuhlen M., Vogelsberger M., Angulo R., 2012, Physics of the Dark Universe, 1, 50
dc.relation.referencesLorenzo-Gonz´alez M., Calvo-Iglesias E., ´ Avarez-Fern´andez I., 2022, Ciencia, T´ ecnica y Mainstreaming Social, pp 33–42
dc.relation.referencesNain A. K., 2022, Denoising Diffusion Probabilistic Mod els, https://keras.io/examples/generative/denoising_ diffusion_probabilistic_models/
dc.relation.referencesNtampaka M., et al., 2021, The Role of Machine Learning in the Next Decade of Cosmology (arXiv:1902.10159)
dc.relation.referencesPlanck Collaboration Aghanim N., Akrami Y., Ashdown M., Au mont J., et al., 2020, Planck 2018 results. VI. Cosmological parameters, https://www.aanda.org/articles/aa/abs/2020/ 09/aa33910-18/aa33910-18.html
dc.relation.referencesRezende D., Mohamed S., 2015, in International conference on machine learning. pp 1530–1538
dc.relation.referencesRombach R., Blattmann A., Lorenz D., Esser P., Ommer B., 2022, High-Resolution Image Synthesis with Latent Diffusion Models (arXiv:2112.10752)
dc.relation.referencesRouhiainen A., 2024, Cosmology at the Field Level with Proba bilistic Machine Learning (arXiv:2402.07694)
dc.relation.referencesSohl-Dickstein J., Weiss E. A., Maheswaranathan N., Ganguli S., 2015, Deep Unsupervised Learning using Nonequilibrium Ther modynamics (arXiv:1503.03585)
dc.relation.referencesSong J., Meng C., Ermon S., 2021, in International Confer ence on Learning Representations. https://openreview.net/ forum?id=St1giarCHLP
dc.relation.referencesSong J., Meng C., Ermon S., 2022, Denoising Diffusion Implicit Models (arXiv:2010.02502) Diffusion Implicit Models (arXiv:2304.03322)
dc.relation.referencesTassev S., Zaldarriaga M., Eisenstein D. J., 2013, Journal of Cos mology and Astroparticle Physics, 2013, 036–036
dc.relation.referencesValogiannis G., Dvorkin C., 2022, Physical Review D, 106
dc.relation.referencesVerbin D., Hedman P., Mildenhall B., Zickler T., Barron J. T., Srinivasan P. P., 2021, Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields (arXiv:2112.03907)
dc.relation.referencesWinther H. A., 2017, MG-PICOLA-PUBLIC, https://github. com/HAWinther/MG-PICOLA-PUBLIC
dc.relation.referencesYang R., Srivastava P., Mandt S., 2023, Entropy, 25, 1469
dc.relation.referencesYankelevich V., Porciani C., 2018, Monthly Notices of the Royal Astronomical Society, 483, 2078–2099
dc.relation.referencesZhang G., Ji J., Zhang Y., Yu M., Jaakkola T., Chang S., 2023, Towards Coherent Image Inpainting Using Denoising Diffusion Implicit Models (arXiv:2304.03322)
dc.rightsAtribución-NoComercial-CompartirIgual 4.0 Internacional
dc.rights.accessrightsinfo:eu-repo/semantics/openAccess
dc.rights.accessrightshttp:/purl.org/coar/access_right/c_abf2/
dc.rights.localAcceso abierto
dc.rights.urihttp://creativecommons.org/licenses/by-nc-sa/4.0/
dc.subjectDeep Learning
dc.subjectModelos generativos
dc.subjectAstrofísica
dc.subjectSimulaciones de N-cuerpos
dc.subjectGravedad modificada
dc.subject.keywordsDeep Learning
dc.subject.keywordsGenerative models
dc.subject.keywordsAstrophysics
dc.subject.keywordsN-body simulations
dc.subject.keywordsModified gravity
dc.titleModelos de difusión probabilísticos como emuladores en simulaciones de gravedad modificada
dc.title.translatedProbabilistic diffusion models as emulators in modified gravity simulations

Archivos

Bloque original
Mostrando 1 - 1 de 1
No hay miniatura disponible
Nombre:
Trabajo de grado.pdf
Tamaño:
2.74 MB
Formato:
Adobe Portable Document Format