English
Enhancing sampling with learned transport maps
Multi angle
Auteurs : Gabrié, Marylou (Auteur de la Conférence)
CIRM (Editeur )
62F15 - Bayesian inference
68T99 - None of the above but in this section
82B80 - Numerical methods in equilibrium statistical mechanics
Ressources complémentaires :
https://cloud.enpc.fr/s/kz6NipeHBDQ7X8K
CIRM (Editeur )
Loading the player...
|
Résumé :
Deep generative models parametrize very flexible families of distributions able to fit complicated datasets of images or text. These models provide independent samples from complex high-distributions at negligible costs. On the other hand, sampling exactly a target distribution, such as Boltzmann distributions and Bayesian posteriors is typically challenging: either because of dimensionality, multi-modality, ill-conditioning or a combination of the previous. In this talk, I will review recent works trying to enhance traditional inference and sampling algorithms with learning. I will present in particular flowMC, an adaptive MCMC with Normalizing Flows along with first applications and remaining challenges.
Keywords : generative models; sampling; MCMC
Codes MSC : Keywords : generative models; sampling; MCMC
62F15 - Bayesian inference
68T99 - None of the above but in this section
82B80 - Numerical methods in equilibrium statistical mechanics
Ressources complémentaires :
https://cloud.enpc.fr/s/kz6NipeHBDQ7X8K
|
Informations sur la Vidéo
Réalisateur : Petit, JeanLangue : Anglais Date de publication : 14/04/2023 Date de captation : 03/04/2023 Sous collection : Research talks arXiv category : Machine Learning Domaine : Numerical Analysis & Scientific Computing ; Probability & Statistics Format : MP4 (.mp4) - HD Durée : 00:38:41 Audience : Researchers ; Graduate Students ; Doctoral Students, Post-Doctoral Students Download : https://videos.cirm-math.fr/2023-04-03_gabrie.mp4 
|
Informations sur la Rencontre
Nom de la rencontre : Analysis and simulations of metastable systems / Analyse et simulation de systèmes métastablesOrganisateurs de la rencontre : Landim, Claudio ; Lelièvre, Tony ; Bianchi, Alessandra Dates : 03/04/2023 - 07/04/2023 Année de la rencontre : 2023 URL Congrès : https://conferences.cirm-math.fr/2742.html
Données de citation
DOI : 10.24350/CIRM.V.20027903Citer cette vidéo: Gabrié, Marylou (2023). Enhancing sampling with learned transport maps. CIRM. Audiovisual resource. doi:10.24350/CIRM.V.20027903 URI : http://dx.doi.org/10.24350/CIRM.V.20027903 |
Voir aussi
- [Multi angle] Metastability of Glauber dynamics with inhomogeneous coupling disorder / Auteur de la Conférence Pulvirenti, Elena.
- [Multi angle] Too many frogs cannot fall sleep / Auteur de la Conférence Gaudillière, Alexandre.
- [Multi angle] Rare transitions in noisy heteroclinic networks / Auteur de la Conférence Bakhtin, Yuri.
Bibliographie
- GABRIÉ, Marylou, ROTSKOFF, Grant M., et VANDEN-EIJNDEN, Eric. Adaptive Monte Carlo augmented with normalizing flows. Proceedings of the National Academy of Sciences, 2022, vol. 119, no 10, p. e2109420119. - https://doi.org/10.1073/pnas.2109420119
- SAMSONOV, Sergey, LAGUTIN, Evgeny, GABRIÉ, Marylou, et al. Local-Global MCMC kernels: the best of both worlds. Advances in Neural Information Processing Systems, 2022, vol. 35, p. 5178-5193. - https://proceedings.neurips.cc/paper_files/paper/2022/hash/21c86d5b10cdc28664ccdadf0a29065a-Abstract-Conference.html
- WONG, Kaze WK, GABRIÉ, Marylou, et FOREMAN-MACKEY, Daniel. flowMC: Normalizing-flow enhanced sampling package for probabilistic inference in Jax. Journal of Open Source Software, 8(83), 5021, 2023. - https://doi.org/10.21105/joss.05021
- GRENIOUX, Louis, DURMUS, Alain, MOULINES, Éric, et al. On Sampling with Approximate Transport Maps. arXiv preprint arXiv:2302.04763, 2023. - https://doi.org/10.48550/arXiv.2302.04763
