Mar 26, 2025 |
Our NSF Postdoctoral Fellow, Allen Alvarez Loya, has been selected as the next Information Science & Technology (ISTI) Postdoctoral Fellows at LANL. He will formally join LANL soon (although he has already been at LANL for two and half years). His lead mentor will be Dan Serino. Congratulations, Allen! |
Mar 17, 2025 |
Qi Tang will serve the organizing committee for the next MFEM community workshop. The workshop is scheduled to be held in Portland in Sep 2025. To learn more about past workshops, click here. |
Mar 12, 2025 |
Qi Tang is selected to be an invited speaker in the upcoming NAPAC 25, the largest domestic conference for accelerator science and technology. He will talk about "Symplectic Neural Network Surrogate Models for Applications to Beam Dynamics." See here for more details. |
Mar 3, 2025 |
A new paper has been submitted on learning stiff dynamics. We build several structures strongly into neural ODE and leverage the advanced time stepping techinque of exponential integrators. The lead author is Allen Alvarez Loya, who is an NSF postdoc in our team. See arXiv. |
Jan 22, 2025 |
A new paper has been accepted in ICLR. We consider graph neural networks and propose a new graph representation of geometric data with 3D coordinates (e.g., molecules). The lead authors are Shih-Hsin Wang, who interned with me last summer, and Prof. Bao Wang from Utah. More details can be found here. This ICLR paper has been selected as an oral presentation (top 1.8%). |
Jan 19, 2025 |
Qi Tang is serving the scientific organizing committee of a two-week LANL workshop focused on multi-physics algorithms beyond exascale computing. The workshop will bring together leading experts to discuss advancements in the field and explore future directions. For more details, click here. |
Jan 6, 2025 |
GT PhD student Alejo Ballester participated in the ICERM program on computational learning for model reduction. We thank ICERM and its organizers for their generous travel support, which made his attendance possible. |
Dec 19, 2024 |
Our team's project has been selected by by Georgia Tech's Open Source Program Office (OSPO). Our goal is to release our Newton-based Grad-Shafranov solver as an open-source project. Future enhancements include MPI support and advanced mesh adaptivity. We currently have an immediate opening for a RA position, ideal for undergraduate or master's students. If you're interested, please don't hesitate to reach out. We gratefully acknowledge the generous support from OSPO and the Sloan Foundation. |
Dec 9, 2024 |
Our team has a project selected at the Computational Physics Student Summer Workshop at LANL next summer. The project, titled "Kinetic Plasma Simulations Using Super-Particles," builds upon the work of Ph.D. student Mandela Quashie. If you are interested in working with us and participating the workshop as a student fellow, please apply by January 15th (US citizen only). |
Dec 4, 2024 |
A new inter-lab LDRD project has been selected for funding. We will collaborate with LANL (PI: Chengkun Huang), SNL (PI: Khachik Sargsyan), and PNNL (PI: Ben McDonald). The project focuses on developing data-driven approaches, including scientific machine learning and reduced-order modeling, for multi-probe radiography systems. A funded PhD position at Georgia Tech will be available as part of this project. |
Nov 22, 2024 |
A new paper has been submitted on learning generalized diffusions using a variational framework. Our network builds in an energy dissipation law strongly into the learning framework. This is a joint work with applied mathematicians from Illinois Tech and UCR, and it was primarily supported by our DOE ASCR SciML project. See arXiv. |
Oct 16, 2024 |
A new paper has been submitted on a structure-preserving discontinuous Galerkin scheme for the Cahn-Hilliard equation. We designed and developed the new scheme using a combination of compatible finite elements, fully implicit time stepping, preconditioning, and time adaptivity. This is a joint work with Golo Wimmer and Ben Southworth at LANL. See arXiv. |
Sep 24, 2024 |
A new paper has been submitted on runaway electrons through wave-wave and wave-particle interactions. Using LANL's flagship VPIC code, We conducted large-scale Particle-In-Cell simulations to discover new physics related slow-X modes. This is a joint work with several physicists at LANL. See arXiv. |
Aug 29, 2024 |
Our postdoc, Dan Serino, has accepted an offer to become a Staff Scientist at T-5. He will continue as a collaborator of our team, with his research focusing on machine learning, applied math, and computational science at LANL. Congratulations, Dan! |
Aug 20, 2024 |
Qi Tang has started a new position as a tenure-track assistant professor in the School of Computational Science and Engineering (CSE) at the Georgia Institute of Technology. At Georgia Tech, our team will continously explore high-performance computing, scientific computing and simulation, and AI/ML, three core research areas at CSE. |
Aug 16, 2024 |
After nearly six fruitful years at Los Alamos National Laboratory, our team will be leaving T-5 and LANL. It has been a remarkable journey collaborating with some of the world’s best computational scientists, mathematicians, and physicists at Department of Energy national labs. We especially thank our collaborators from LANL, LLNL, SNL, and ANL for many helpful discussions. |
Aug 8, 2024 |
Our postdoc, Xuping Xie, has started a new position as a tenure-track assistant professor in Department of Mathematics and Statistics and School of Data Science at Old Dominion University. Congratulations! |
Jul 3, 2024 |
A new paper has been submitted on an adaptive Newton-based method for the free-boundary Grad-Shafranov equation. This solver provides an accurate low-beta initial condition for whole device modeling (WDM) in magnetic confinement fusion. See arXiv. Our postdoc, Dan Serino, is the lead author. |
Jun 17, 2024 |
A new paper has been submitted on building a physics-assisted latent dynamics model for collisional-radiative models. This is a critical component in integrated modeling for fusion WDM. See OSTI. Our postdoc, Xuping Xie, is the lead author. |
Feb 24, 2024 |
A new paper has been submitted on building Fenichel normal form strongly into neural ODE. This is the first realization of Fenichel normal form in machine learning, opening a new path for studying multiscale dynamical systems. See arXiv. Our postdocs, Dan Serino and Allen Alvarez-Loya, are the lead authors. |