Master Thesis Online
Master Thesis : Diffusion models : seek of information and structure in latent space
Maziane, Yassine 
Promotor(s) :
Louppe, Gilles
Date of defense : 4-Sep-2023/5-Sep-2023 • Permalink : http://hdl.handle.net/2268.2/18351
TFE_to_deliver.pdfDetails
| Title : | Master Thesis : Diffusion models : seek of information and structure in latent space |
| Author : | Maziane, Yassine
|
| Date of defense : | 4-Sep-2023/5-Sep-2023 |
| Advisor(s) : | Louppe, Gilles
|
| Committee's member(s) : | Wehenkel, Louis
Geurts, Pierre
|
| Language : | English |
| Number of pages : | 137 |
| Keywords : | [en] Generative AI [en] Diffusion models [en] Latent spaces |
| Discipline(s) : | Engineering, computing & technology > Computer science |
| Target public : | Researchers Professionals of domain Student |
| Institution(s) : | Université de Liège, Liège, Belgique |
| Degree: | Master : ingénieur civil en science des données, à finalité spécialisée |
| Faculty: | Master thesis of the Faculté des Sciences appliquées |
Abstract
[en] This thesis investigates diffusion models, a subset of generative modeling, focusing on when and how they make feature choices during data generation. Contrary to prevailing beliefs of instant choices, experiments reveal that these decisions evolve gradually over denoising steps. Feature preservation and destruction in the forward process are guided by noise variance schedules, while crucial feature choices occur symmetrically in the diffusion process.
Cite this master thesis
All documents available on MatheO are protected by copyright and subject to the usual rules for fair use.
The University of Liège does not guarantee the scientific quality of these students' works or the accuracy of all the information they contain.
The University of Liège does not guarantee the scientific quality of these students' works or the accuracy of all the information they contain.