Scholar
Yifan Chen
Google Scholar ID: GNiinUoAAAAJ
UCLA Mathematics
Citations & Impact
Citations
760
H-index
14
i10-index
14
Publications
20
Co-authors
40
list available
Contact
Publications
Resume (English only)
Academic Achievements
- Demonstrated 'delocalization of bias' in unadjusted Langevin algorithms, leading to nearly dimension-independent bias scaling under certain conditions.
- Developed new derivative-free and derivative-based affine-invariant ensemble samplers that improve upon the emcee package, especially in high dimensions.
- Introduced affine-invariant ensemble HMC achieving state-of-the-art dimensional scaling while maintaining affine invariance and embarrassingly parallel structure.
- Established a design framework for gradient flows in sampling, focusing on Fisher-Rao geometry, and developed a Kalman-filter-type Gaussian mixture approximation for large-scale Bayesian inverse problems.
- Proposed principled designs for noise and interpolation schedules in generative modeling of multiscale scientific data, introducing 'optimal averaged Lipschitzness' for improved numerical efficiency on ill-conditioned distributions.
