Learning dynamical systems
Learning complex dynamical systems through representation learning and generative modeling.
Representations · Generative models · State-space modelsPh.D. candidate · Columbia University
Yongquan Qu.
Dynamics, inference
& uncertainty.
I am interested in machine learning and probabilistic inference for complex dynamical systems, with applications in weather and climate.
Currently atColumbia UniversityNYU Courant
01 / About
About me.
I am a Ph.D. candidate at Columbia University, advised by Professor Pierre Gentine. I’m currently visiting NYU Courant as a research affiliate, working with Professor Laure Zanna. I am also a graduate research assistant at LEAP and affiliated with the M2LInES project.
Internship experience
Pasteur Labs ↗
Simulation Intelligence Intern / Resident
I worked on scientific machine learning research and development, first as an intern and later as a resident.
Previously · HKUST & Lanzhou University
Before coming to Columbia, I obtained my M.Phil. in Atmospheric Environmental Science at HKUST, supervised by Xiaoming Shi, working on differentiable programming and transfer learning for hybrid ML-physics models. I received my B.Sc. in Mathematics and Applied Mathematics from Lanzhou University, where I worked on compressed sensing and sparse representation in Yu-Mei Huang’s group.
02 / Research interests
Questions I’m interested in.
Inference & uncertainty
Inferring hidden states and uncertainty from sparse observations.
Bayesian inference · Generative modelsScientific machine learning
Combining machine learning and physics for weather and climate.
Differentiable programming · Forecasting03 / Selected work
Recent research.
Structural Blindness in Latent Data Assimilation: Representation Geometry Misleads Sensor Design
Yongquan Qu, Pierre Gentine
Video Diffusion for Satellite-based High-Dynamical-Fidelity Precipitation (HiDFiP) Field Generation
Runze Li, Yan Xia, Yongquan Qu, Dongwei Fu, Clement Guilloteau, Judy Hoffman, Stephan Mandt, Pierre Gentine, Efi Foufoula-Georgiou
Strictly Constrained Generative Modeling via Split Augmented Langevin Sampling
Matthieu Blanke, Yongquan Qu, Sara Shamekh, Pierre Gentine
Physically consistent global atmospheric data assimilation with machine learning in latent space
Hang Fan, Ben Fei, Pierre Gentine, Yi Xiao, Kun Chen, Yubao Liu, Yongquan Qu, Fenghua Ling, Lei Bai
04 / Updates
Notes & news.
-
(Sep. 2026) Our paper Structural Blindness in Latent Data Assimilation: Representation Geometry Misleads Sensor Design was accepted to NeurIPS 2026.
-
(2026) Our paper Data-driven subgrid-scale turbulence parameterization: methods for improving model efficiency was published in Machine Learning: Earth. paper
-
(Aug. 2026) Our preprint Video Diffusion for Satellite-based High-Dynamical-Fidelity Precipitation (HiDFiP) Field Generation is available on arXiv.
-
(Jan. 2026) Our paper Strictly Constrained Generative Modeling via Split Augmented Langevin Sampling was accepted to ICLR 2026. See you in Rio de Janeiro!
-
(2026) Our paper Physically consistent global atmospheric data assimilation with machine learning in latent space was published in Science Advances. paper
-
(2026) Our paper LEX v1.6.0: a new large-eddy simulation model in JAX with GPU acceleration and automatic differentiation was published in Geoscientific Model Development. paper
-
(Oct. 2025) Our preprint on Incorporating Multivariate Consistency in ML-Based Weather Forecasting with Latent-space Constraints is available on arXiv.
-
(Aug. 2025) Our preprint PnP-DA: Towards Principled Plug-and-Play Integration of Variational Data Assimilation and Generative Models is available on arXiv.
-
(Dec. 2024) Our paper ChaosBench: A Multi-Channel, Physics-Based Benchmark for Subseasonal-to-Seasonal Climate Prediction appeared in NeurIPS 2024.
-
(Jul. 2024) I joined Pasteur Labs as a Simulation Intelligence Intern, working on research and development in scientific machine learning.
-
(Apr. 2024) Our paper Deep Generative Data Assimilation in Multimodal Setting was accepted at the CVPR 2024 EarthVision Workshop and received the Best Student Paper Award.
-
(Mar. 2024) Our proposed ICML 2024 workshop Machine Learning for Earth System Modeling: Accelerating Pathways to Impact was accepted.
-
(Mar. 2024) Our paper Joint Parameter and Parameterization Inference with Uncertainty Quantification through Differentiable Programming was accepted at the ICLR 2024 Workshop on AI4DifferentialEquations in Science.
Let’s connect