Philip Greengard

About Me

I am a Postdoctoral Research Scholar at Columbia University in Andrew Gelman's applied statistics group. I work on both applied statistics and developing efficient numerical methods for statistical inference. I got a PhD in Applied Mathematics in 2019 from Yale where I worked with Vladimir Rokhlin. Here's a short cv.

Selected Publications and Pre-prints

Applied and computational statistics

• An improved BISG for inferring race from surname and geolocation (with A. Gelman)
• Equispaced Fourier representations for efficient Gaussian process regression from a billion data points (with A. H. Barnett, M. Rachh)
• Efficient Fourier representations of families of Gaussian processes
• Fast methods for posterior inference of two-group normal-normal models (with J. Hoskins, C. Margossian, J. Gabry, A. Gelman, A. Vehtari)
• Efficient reduced-rank methods for Gaussian processes with eigenfunction expansions (with M. O'Neil)

Numerical analysis

• Uniform approximation of common Gaussian process kernels using equispaced Fourier grids (with A. Barnett, M. Rachh)
• Generalized prolate spheroidal functions: algorithms and analysis
• An algorithm for the evaluation of the incomplete gamma function (with V. Rokhlin)

Efficient deep learning

• LQ-LoRA: Low-rank Plus Quantized Matrix Decomposition for Efficient Language Model Finetuning (with H. Guo, E. P. Xing, Y. Kim)
• Learning to grow pretrained models for efficient transformer training (with P. Wang, R. Panda, L. T. Hennigen, L. Karlinsky, R. Feris, D. D. Cox, Z. Wang, Y. Kim)
• Federated learning as variational inference: A scalable expectation propagation approach (with H. Guo, H. Wang, A. Gelman, E. Xing, Y. Kim)

Contact Me

pg2118 at columbia.edu