Youzuo Lin

📣 Competition Launch Alert! Geophysical Waveform Inversion hosted by @Yale & @UNC 🎯 Develop physics-guided ML models to solve full-waveform inversion problems 💰 $50,000 Prize Pool ⏰ Entry Deadline: June 23, 2025 kaggle.com/competitions/w…

The Reaction Says it All 🙌

It is interesting to see how the focus of applications of neural networks in seismology has changed over time. #DeepLearningSeismology #AIforEarth #Seismology #Earthquakes #ExplorationSeismology #MachineLearning

Please check out our recent collaborative work in JGR (co-supervised with Dr. Yuozuo Lin @YouzuoLin1). We develop a Spatiotemporal Graph Neural Network (STGNN) for estimating earthquake locations and magnitudes. agupubs.onlinelibrary.wiley.com/doi/full/10.10…

Very excited to share our OpenFWI was accepted by #NeurIPS. Great job Chengyuan, Shihang, and everyone contributing to this effort! OpenFWI: arxiv.org/abs/2111.02926

I am extremely happy to share that the #ICML2022 submission by us has been accepted for a short presentation. We show an interesting near-linear relationship after applying the integral transform. Great job, Yinan & the team! Here is our work: arxiv.org/abs/2204.13731

It wasn't dinosaurs' best day ever. Watch LANL scientist Cathy Plesko talk about #asteroid impacts and how to deflect them on the "Dinosaur Apocalypse" episode of NOVA on @PBS on May 11. pbs.org/wgbh/nova/seri…

Excited to share our newest review paper on the topic of Physics-Informed Data-Driven Seismic Inversion. Here is the work: essoar.org/pdfjs/10.1002/…

@Tieyuan_Zhu Thanks for the reply! 😀 Like what you said, the quantity and quality of the training set are equally important. Our method learns the physics knowledge (i.e. physics perception) and uses it to make sure the synthetic velocity maps generated would be physically meaningful.

@YouzuoLin1 This idea sounds crazy! How realistic of these maps or how are they honoring the geology? Perhaps these questions are not important for training. We just need more training samples. Neat!

You don't have sufficient seismic velocity maps to train your data-driven inversion models? Hey, our method turns natural images into velocity maps. Check it out below: Paper: ieeexplore.ieee.org/document/95566…

Look forward to the upcoming SEG Workshop on Data Analytics and Machine Learning for Exploration and Production! Here is the agenda to download: seg.org/Portals/0/Open…

@meta_Innovation This seems like very nice functionalities and also a user-friendly GUI. 👍

Launching this month on the #metaKinetic platform, "Seismic Inversion" simulation combines seismic forward modeling and inversion to ensure the users learn this method interactively through cause and effect. Check it out 👇 #oilandgas #geophysics #seismic #interactivelearning

Excited to share our ICLR work "Unsupervised Learning of Full-Waveform Inversion: Connecting CNN and Partial Differential in a Loop", as a poster presentation this year! Nice job, Peng and Xitong! 👍 Paper: openreview.net/pdf?id=izvwgBi…

My first Twitter Message!!