Touring sampling with pushforward maps
This work provides a theoretical framework for practitioners to navigate and improve sampling methods in generative modeling, though it is incremental as it reviews and links existing techniques rather than introducing new ones.
The paper organizes and reviews various sampling methods in generative modeling to help practitioners apply machine learning techniques to their specific problems, revealing connections between existing approaches to address challenges like long inference times and lack of diversity in diffusion models.
The number of sampling methods could be daunting for a practitioner looking to cast powerful machine learning methods to their specific problem. This paper takes a theoretical stance to review and organize many sampling approaches in the ``generative modeling'' setting, where one wants to generate new data that are similar to some training examples. By revealing links between existing methods, it might prove useful to overcome some of the current challenges in sampling with diffusion models, such as long inference time due to diffusion simulation, or the lack of diversity in generated samples.