Peter Yichen Chen
189 posts
Peter Yichen Chen
@peterchencyc
UBC PhysAI Lab – Hiring postdocs, PhD students, and research interns
Joined November 2020
- #SIGGRAPH2024 Neural network and Monte Carlo are two ways to solve your PDEs without any mesh or grid. @jn_pranav combines them to push the boundary of grid-free simulations. We hope this will inspire more work combining these two approaches! See more: pranav-jain.github.io/projects/nmcfs/
- People say there’s a huge gap between simulation and reality—especially when you run the simulation for a long time. That’s generally true… but we’re excited to share that we’ve taken a solid step toward closing that gap. We can now accurately simulate a robot folding boxes in
00:00 - #SIGGRAPHAsia2023 #NeuralPDE Neural Stress Field is a reduced-order modeling framework for solid, fluid, and fracture mechanics. After training, it enables 10X speedup for simulations of tearing and other damage phenomena to everyday materials such as bread! Come to the NEURAL
00:00 - #BestPaperAward #SIGGRAPH2025 One neural PDE model, hundreds of shapes — simulated at lightning speed. 🚀 Introducing Shape Space Spectra: first eigenanalysis across shapes. Come see ChangYue’s talk today 👉 changy1506.github.io
00:00 - #BestPaperAwardHonorableMention #SIGGRAPH2025 One neural PDE model, hundreds of cuts and discontinuities — simulated in real time. 🚀 Introducing WindLifter: lifting the winding number fields lets us capture discontinuities in neural fields. Come see ChangYue’s talk today 👉
00:00 - Excited to share our open-source project on building world models through differentiable robot-object interactions! Explore it here: github.com/MediosZ/WarpDi…. Huge thanks to the @nvidia Warp team (@eric_heiden @milesmacklin) for the easy-to-use infrastructure.
00:00 - #SIGGRAPH2024 #HybridNeuralPDE Nvidia Warp is an excellent tool for a gazillion things! For me personally, it’s a key building block for integrating classic numerical PDE solvers with AI. Learn more at our SIGGRAPH course tomorrow (Sunday) morning!Discover applications of GPU-accelerated data generation and spatial computing with NVIDIA Warp -- a cutting-edge #Python framework for simulation, #AI, #robotics, and #machinelearning. Join us for hands-on projects with @MIT, @UCLA, and our teams. ➡️ nvda.ws/4bJwHJP
- (Surprise?) Our neural-PDE solver is more accurate than both classic FEM and CFD. The caveat: it’s slower. Come to our session and find out why! #ICML2023 #INSRPDE #AI4SCIENCE Tue 2-3:30 Exhibit Hall 1 #708 Project page: cs.columbia.edu/cg/INSR-PDE/
- #SIGGRAPHAsia2023 #NeuralPDE LiCROM is a geometry-agnostic ML framework for PDEs. After training, it enables 50X speedup while enabling modifying the geometry and even swapping the entire geometry! Check out our live-screen, real-time capture below📷 Come to the DEFORMABLE
00:00 - (Surprise?) Our neural-PDE solver generalizes to extremely out-of-distribution geometries, boundary conditions, and even multi-physics. Hint: we learned a constitutive model. #ICML2023 #AI4SCIENCE #NCLAW #GNN Tue 11-1:30 Exhibit Hall 1 #813 Project page: sites.google.com/view/nclaw
- arxiv.org/abs/2311.12198 Another 3D Gaussian splatting paper? Like the original Gaussian splatting paper, this one feels like a game changer tho. For the first time, "what you see is what you simulate" doesn't seem too far off. Kudos to the UCLA team! math.ucla.edu/aivc/
- My amazing @MIT_CSAIL colleague @Chao_Liu_ is launching the PRIME Robotics Lab at @UBC. Join him and the dynamic UBC AI ecosystem to shape the future of embodied AI! Learn more: chaoliu.tech Retweets appreciated!
