Research Interests
- AI for Science and Physics: neural operators, PDE-based modeling, operator learning
- Scientific ML for limited data and cross-resolution generalization
- Physics-constrained generative models, diffusion models, and sampling
- Stochastic optimization: simulated annealing, RL-guided sampling, and scheduling
Representative Topics
- Physics-informed operator learning and active data acquisition
- Cross-resolution learning and discretization robustness
- Hybrid optimization dynamics and quasi-equilibrium constraints
- Generative backmapping and molecular structure generation
Publications
Papers are divided into conferences, journals, and workshops. Items are ordered newest to oldest within each category.
Peer-Reviewed Conferences
-
IAGA: Identity-Aware Gaussian Approximation for Efficient 3D Molecular Generation
International Conference on Learning Representations (ICLR 2026), Accepted
J. Qu, W. Gao, R. Xu, Y. Liu -
Velocity-Inferred Hamiltonian Networks: Symplectic Dynamics from Position-Only Observations
NYSDS 2025, SIAM Proceedings
R. Xu, C. Yu, Z. Wu, S. Wang, L. Chen, G. Kementzidis, H. Wang, Y. Deng -
Discretization-invariance? On the Discretization Mismatch Errors in Neural Operators
International Conference on Learning Representations (ICLR 2025) (ICLR Page)
W. Gao, R. Xu, Y. Deng, Y. Liu -
Kolmogorov–Arnold Representation for Symplectic Learning: Advancing Hamiltonian Neural Networks
International Joint Conference on Neural Networks (IJCNN 2025)
Z. Wu, R. Xu†, L. Chen, G. Kementzidis, S. Wang, Y. Deng
†Co-first author; project lead. -
Boundary-Informed Method of Lines for Physics-Informed Neural Networks
NYSDS 2025, SIAM Proceedings
M. Cederholm, S. Wang, H. Wang, R. Xu, Y. Deng
Co-corresponding author with Y. Deng. -
Physics-Informed Active Learning via Functional Simulated Annealing for Neural Operator
NYSDS 2025, SIAM Proceedings
A. Ding, S. Wang, H. Wang, R. Xu, Y. Deng
Co-corresponding author with Y. Deng.
Peer-Reviewed Journals
-
Multistep Generative Backmapping of Coarse-Grained Structures
Computer Physics Communications, 327, 110286, 2026 (Elsevier Share Link)
G. Kementzidis, E. Wong, J. Nicholson, R. Xu, Y. Deng -
Dynamic Schwartz-Fourier Neural Operator for Enhanced Expressive Power
Transactions on Machine Learning Research (TMLR), 2025
W. Gao, J. Luo, R. Xu, Y. Liu -
Coordinate Transform Fourier Neural Operators for Symmetries in Physical Modeling
Transactions on Machine Learning Research (TMLR), 2024
W. Gao, R. Xu, H. Wang, Y. Liu
Peer-Reviewed Workshops
-
APOD: Adaptive PDE-observation diffusion for physics-constrained sampling
ICML 2025 Workshop on Assessing World Models (Workshop Website) (ICML Page)
R. Xu, H. Wang, G. Kementzidis, C. Si, Y. Deng -
RL-QESA: Reinforcement-Learning Quasi-Equilibrium Simulated Annealing
2nd AI for Math Workshop @ ICML 2025
R. Xu, K. Li, H. Wang, G. Kementzidis, W. Zhu, Y. Deng -
An Iterative Framework for Generative Backmapping of Coarse-Grained Proteins
ICML 2025 GenBio Workshop
G. Kementzidis, E. Wong, J. Nicholson, R. Xu, Y. Deng
In Preparation / Ongoing
- JEPA for PDEs and Graph-Structured Scientific Data, 2026.
- ORACLE: LLM-Guided Edits + Simulated Annealing for Structure-Based Drug Design, 2026.
- PartialObs–PDEBench: Partial-Observation Benchmark for PDE Learning, ongoing.