Pu Ren

I am a Postdoc Fellow in AI & Learning Systems Group at Lawrence Berkeley National Lab, where I am working with Prof. Michael Mahoney. I received my Ph.D. at Northeastern University in July 2022, where I worked with Prof. Hao Sun and Prof. Ryan Wang. I was also fortunate to collaborate with Prof. Jerome F. Hajjar, Prof. Yang Liu, and Prof. Jian-Xun Wang. I am broadly interested in AI methods for scientific problems. I work on both the methodology and application sides of scientific machine learning. Specifically, my research focuses on soft- and hard-constrained machine learning, generative modeling, and large pretrained models, with applications spanning fluid dynamics, climate science, geophysics, earthquake engineering, etc. More recently, I have been exploring weight-space analysis and its role in scientific machine learning, particularly through loss landscape analyses of constrained models and eigen-spectral analyses of large pretrained models. My ongoing projects aim to establish theoretical and computational foundations that connect these analyses to the generalization and transferability of scientific foundation models. Email: pren AT lbl DOT gov

Recent News

• [07/2024] We are organizing a workshop at NeurIPS 2024 on Foundation Models for Science. We welcome submissions on scientific foundation models, AI for Science, and Scientific Machine Learning (SciML). See more details using this Link.

• [07/2024] Our work on conditional generative modeling for earthquakes is online now. (arXiv)

• [06/2024] WaveCastNet is online now. We proposed an AI-enabled wavefield forecasting framework for earthquake early warning. (arXiv)

• [05/2024] I presented our recent work on conditional VAE models for ground motion modeling at Seismological Society of America 2024.

• [05/2024] Our work on data-efficient pretraining and in-context learning for SciML has been accepted by ICLR 2024 Workshop on AI4Differential Equations In Science. (arXiv)

• [04/2024] Bailey has put our PhD work online. We investigated the physics-informed ML for seismic response modeling. (arXiv)

• [01/2024] Our work on generative models for time series data has been accepted by ICLR 2024. (arXiv)

• [12/2023] I presented our recent work on Deep Generative Models for Earthquake Ground Motion Simulation at AGU 2023.

• [12/2023] Our work on ground‐motion spectra clustering and selection has been published on Earthquake Engineering & Structural Dynamics. (DOI) (arXiv)

• [11/2023] Our work on physics-informed deep learning for seismic wave propagation has been published in the Computer Physics Communications. (DOI) (arXiv)

• [08/2023] Our work on physics-informed super-resolution for spatiotemporal data has been published in the Journal of Computational Physics. (DOI) (arXiv)

• [06/2023] Our work SuperBench is online now. We build a super-resolution benchmark for scientific machine learning! (arXiv)

• [06/2023] Our work on encoding physics to learn reaction-diffusion processes has been accepted by Nature Machine Intelligence. (DOI) (arXiv)

• [04/2023] Our work on seismic inversion using PINNs is online now. (arXiv)

Selected Awards

• [05/2021] Second Place of 2021 EMI Dynamics Committee Student Paper Competition (Link)

• [05/2020] Third Place of 2020 EMI SHM&C Committee Student Paper Competition (Link)

• [09/2019] Vilas Mujumdar Fellowship

• [10/2018] China National Scholarship

• [04/2015] Finalist of Mathematical Contest in Modeling (top 0.5%)