Jiajun He jh2383@cam.ac.uk
China
Engineering, Darwin College
PhD thesis: TBC
Research interests:
- Probabilistic models and Bayesian machine learning
- Data and model compression
- Relative entropy coding, channel simulation
- Diffusion models
The rise of deep generative models has opened up a new path in data compression. More efficient and flexible coding frameworks have been proposed and studied, leveraging the capabilities of machine learning models. My PhD research will focus on pushing the boundaries in this area by exploring Bayesian and probabilistic methods for data compression. Specifically, I plan to concentrate on relative entropy coding, aiming to create faster and more practical algorithms, and to expand the range of applications for these techniques. Additionally, I have a strong interest in the integration of generative models with relative entropy coding to further enhance compression performance.
Who or what inspired you to pursue your research interests?
The immortality of knowledge gives me a deep sense of satisfaction.