Qilong Pan

PhD Candidate in Statistics and High Performance Computing

About Me

Qilong Pan

I am a PhD candidate in Statistics and High Performance Computing at King Abdullah University of Science and Technology (KAUST). My research focuses on statistical methods and computational models in data science.

My research interests include but are not limited to:

  • 1. Statistical Learning
  • 2. High Performance Computing
  • 3. GPU acceleration
Google Scholar

Publications & Preprints (Highlighted)

Scaled Block Vecchia Approximation for High-Dimensional Gaussian Process Emulation on GPUs

Pan, Qilong and Abdulah, Sameh and Abduljabbar, Mustafa and Ltaief, Hatem and Herten, Andreas and Bode, Mathis and Pratola, Matthew and Fadikar, Arindam and Genton, Marc G and Keyes, David E and Sun, Ying

arXiv:2504.12004

pipeline parallel

We propose a block Vecchia approximation for scalable and efficient Gaussian process computations. Our approach leverages the Vecchia approximation to reduce the computational complexity of Gaussian process computations, allowing for faster and more efficient inference.

Block Vecchia Approximation for Scalable and Efficient Gaussian Process Computations

Qilong Pan, Sameh Abdulah, Marc G. Genton, Ying Sun

Technometrics, 2025

Fast and accurate estimation log-likelihood Accurate estimation Estimation of parameters

We propose a block Vecchia approximation for scalable and efficient Gaussian process computations. Our approach leverages the Vecchia approximation to reduce the computational complexity of Gaussian process computations, allowing for faster and more efficient inference.

GPU-Accelerated Vecchia Approximations of Gaussian Processes for Geospatial Data using Batched Matrix Computations

Qilong Pan, Sameh Abdulah, Marc G. Genton, David E. Keyes, Hatem Ltaief, Ying Sun

ISC High Performance, 2024

Fast and accurate estimation log-likelihood Accurate estimation

We propose a GPU-accelerated Vecchia approximation for scalable and efficient Gaussian process computations. The proposed framework using the KBLAS library is tested on NVIDIA V100/A100/H100 GPUs and achieves 800X to 1300X speedup compared to the ExaGeoStat (exact GPs) implementation.

Education

PhD in Statistics and High Performance Computing

King Abdullah University of Science and Technology, KSA

Sep 2021 - Present

B.S. in Statistics

Wuhan University of Technology, China

Sep 2017 - June 2021

B.A. in English (Dual Degree)

Huazhong University of Science and Technology, China

Jan 2019 - June 2021

Contact