Hi there! I joined Lawrence Berkeley National Laboratory in August 2022 as a postdoc, working under the supervision of Dr Chao Yang. Prior to that, I obtained my PhD from Purdue University, where I was advised by Professor Haizhao Yang. Before my doctoral studies, I earned my MSc degree from the National University of Singapore and my BSc from Sun Yat-sen University.
Research Interest: Scientific machine learning, deep learning algorithm and interdisciplinary application.
Contact: senweiliang [at] lbl [dot] gov
Solving PDEs on unknown manifolds with machine learning
We propose mesh-free deep learning method and theory based on diffusion maps for solving elliptic PDEs on unknown manifolds, identified with point clouds.
S Liang, S Jiang, J Harlim, H Yang, Applied and Computational Harmonic Analysis, Volume 71, 101652, 2024 [PDF Code].
Probing reaction channels via reinforcement learning
We propose deep learning framework to study rare transition including using reinforcement learning to identify reactive regions and employing NN-based PDE solver to approximate the committor function.
S Liang, AN Singh, Y Zhu, DT Limmer, C Yang, Machine Learning: Science and Technology 4 (4) 2023 [PDF, Code].
Machine learning for prediction with missing dynamics
We developed a deep learning method to recover missing dynamics resulting from partial understanding or observation of physical processes and the computational expense of numerical simulations.
J Harlim, S Jiang, S Liang, H Yang, Journal of Computational Physics 428, 109922, 2021 [PDF, Code].
Effective many-body interactions in reduced-dimensionality spaces through neural networks
We introduce a new paradigm to learn the effective Hamiltonian in data-limited scenario.
S Liang, K Kowalski, C Yang and NP Bauman, arXiv:2407.05536 [PDF].
Finite expression method for solving high-dimensional partial differential equations
We introduce a sympolic approach for high dimensional PDE that seeks an approximate PDE solution in the space of functions with finitely many analytic expressions and, hence, this methodology is named the finite expression method (FEX).
This page has been accessed at least times since 10/10/2021, and on average per day.