南洋理工大学PhD Position in Diffusion based generative models and their application to generative data augmentation and uncertainty quantification申请条件要求-申请方
Organizer
Nanyang Technological University
Description
About the Project
Diffusion-based generative models for images such as Ho et al. (2020), Dockhorn et al. (2021) have made significant impact, and nowadays become routinely used in smartphones.
An interesting potential application of such methods for Machine Learning applications is to use them to increase the diversity of existing datasets via generative data augmentation, allowing us to provide better performance, or similar level of performance with much less labelled training data.
Although there have been promising initial results in this area such as Zheng et al (2023), this is not yet completely understood, and there are still significant improvements needed in computational efficiency before such methods can be widely adopted.
We will develop new algorithms that will be numerically evaluated on various machine learning test problems. These methods will improve the consistency and computational scalability of generative data augmentation.
Explainability and uncertainty quantification will be considered as well.
In addition to algorithmic advances, the project will also develop a mathematical theory showing consistency results for our methods.
Applications for an August 2025 start will be considered until the 15th of May, 2025.
References
Zheng, C., Wu, G., & Li, C. (2023). Toward understanding generative data augmentation. Advances in neural information processing systems, 36, 54046-54060.
Ho, J., Jain, A., & Abbeel, P. (2020). Denoising diffusion probabilistic models. Advances in neural information processing systems, 33, 6840-6851.
Dockhorn, T., Vahdat, A., & Kreis, K. (2021). Score-based generative modeling with critically-damped langevin diffusion. arXiv preprint arXiv:2112.07068.
Dombrowski, A. K., Gerken, J. E., Müller, K. R., & Kessel, P. (2023). Diffeomorphic counterfactuals with generative models. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(5), 3257-3274.
Paulin, D., Whalley, P. A., Chada, N. K., & Leimkuhler, B. (2024). Sampling from Bayesian neural network posteriors with symmetric minibatch splitting Langevin dynamics. arXiv preprint arXiv:2410.19780., Accepted for AISTATS 2025.
Nanyang Technological University PhD Position in Diffusion based generative models and their application to generative data augmentation and uncertainty quantification项目有没有奖学金，是不是全奖Phd招生，下面我们一起看一下【大学名称】Phd的奖学金资助情况
Honors
A fully funded NTU Research Scholarship is available for this project,
https://www.ntu.edu.sg/admissions/graduate/financialmatters/scholarships/rss#Content_C006_Col00 .
For international applicants, this is 2700 SGD for the first year, and 3200 SGD later on after passing the required courses and doing an oral presentation about the research plans.
How to register
Professor
扫描二维码
下载申请方APP
下载申请方APP
扫描二维码
下载申请方APP