🚀🎉 Excited to announce 🌟 PyTorch Frame 🌟 - our new open-source initiative in PyTorch! Dive into multi-modal tabular deep learning like never before!
Link:
#PyTorch
#OpenSource
(1/6)
Super excited to share Open Graph Benchmark (OGB)! OGB provides large-scale, diverse graph datasets to catalyze graph ML research. The datasets are easily accessible via OGB Python package with unified evaluation protocols and public leaderboards.
Paper:
Thrilled to share that I've been honored with the prestigious
#KDD2023
Doctoral Dissertation Award! 🏆 Special thanks to my advisor
@jure
for the support and guidance throughout this journey.
Happy to find that the CS224w lecture on "Limitations of Graph Neural Networks" is out on YouTube. I helped making the slides :) The entire course is highly recommended for folks interested in graph ML!
Video:
Course website:
Our paper "How powerful are graph neural networks?" was accepted as ICLR oral! Joint work with Keyulu Xu, Jure Leskovec and Stefanie Jegelka. Check out our paper here:
#iclr2019
Excited to announce that "Open Graph Benchmark: Datasets for ML on Graphs" has been accepted to
#NeurIPS
as a spotlight!
If you are working on Graph ML, please check it out for your next conference submission :)
Website:
Paper:
Excited to share new work, TuneUp!
Many GNN models developed but their training strategies remain simplistic. We present TuneUp to better train your GNN and improve its predictive performance on a wide range of node-/edge-level tasks!
Paper:
(1/N)
Super excited to share our work "Pre-training Graph Neural Networks."
We explore effective strategies for pre-training GNNs, significantly boosting predictive performance across many downstream tasks in two scientific domains (chemistry and biology).
If you are at
#KDD2022
, check out our poster and talk on "Learning Backward Compatible Embeddings"!
We present a setting/solution to ensure backward compatibility of embeddings for downstream models even when the embedding version is updated over time.
@kaidicao
will present!
Excited to share our spotlight paper at
#ICLR2020
"Strategies for Pre-training GNNs." We develop effective strategy/methods for pre-training GNNs and systematically study its effectiveness.
Web site:
Paper:
Check out our PyTorch Frame tech report
We aim to push tabular deep learning to handle complex multi-modal, multi-tabular data, beyond conventional single-tabular data. We do so by seamlessly integrating PyTorch Frame with LLMs and Graph Neural Nets!
🚀🎉 Excited to announce 🌟 PyTorch Frame 🌟 - our new open-source initiative in PyTorch! Dive into multi-modal tabular deep learning like never before!
Link:
#PyTorch
#OpenSource
(1/6)
Genuinely amazed by all the graph models/methods developed on the Open Graph Benchmark (). ~500 leaderboard submissions in 2.5 years, with reproducible code + technical reports. Incredible accuracy improvements across the board (some almost solved).
Excited to see that my molecular property predictor system "Pre-trained OGB-GIN" has just ranked top at the COVID-19 open task hosted by MIT AICures team . Thanks for the amazing effort, and it was exciting to apply my research to this timely problem.
Today is my last day at
@Kumo_ai_team
. I am truly grateful for the opportunities and learning experiences I've had here. Glad to witness the company's hyper-growth and proud to see my research being deployed to many customer use cases.
#ThankYou
#NextChapter
Very excited to share our KDD 2022 paper "Learning Backward Compatibility Embeddings."
We formulate a unique challenge of using embeddings in large industrial organizations and present an effective solution to it.
Paper:
Code:
A nice summary of recent trends in Graph Machine Learning. It also touches upon our Query2box framework for answering complex queries over incomplete knowledge graphs :)
@ren_hongyu
@jure
Excited to give a contributed talk on "ForceNet: A GNN for Large-Scale Quantum Chemistry Calculations” at Deep Learning for Simulation Workshop at ICLR 2021
Paper:
Talk:
Done as part of Open Catalyst Project
Presenting GraphCast: our state-of-the-art AI model delivering 10-day weather forecasts with unprecedented accuracy in under one minute. 🌦️
It can even help predict the potential paths of cyclones further into the future.
Here's how it works. 🧵
Updates from the Open Graph Benchmark Large-Scale Challenge @ KDD Cup 2021.
- Over 500 teams registered.
- Initial test leaderboards released at
- Final test submission deadline on June 8.
Excited to see great progress in large-scale graph ML!
Very horned and excited to give a talk at KDD Deep Learning Day (Aug 24)!
I will present our recent advances in Graph Neural Networks: expressive power, pre-training, and OGB. Thanks for the organizers for kind invitation.
Excited to share my first work in my Ph.D.: "How powerful are graph neural networks?" We provide a theoretical framework for comparing the expressive power of Graph NNs. Great collaboration with Keyulu from MIT and our advisors, Jure and Stefanie! Link:
Check out our 2nd major release of OGB!
We have provided 5 new datasets, including a web-scale gigantic graph (with 100M+ nodes) and heterogeneous graphs.
Excited to announce 2nd major release of Open Grpah Benchmark (OGB), a collection of realistic, large-scale, and diverse benchmark datasets for machine learning with graphs.
The Open Graph Benchmark (OGB) leaderboards are now on Papers with Code!
This provides a comprehensive way to compare models across graph-related tasks and datasets.
Just found the video recording of my talk on Graph Neural Networks at the KDD Deep Learning day available on YouTube!
It's a short 15min summary of the first half of my Ph.D. 🙃 Check it out if you are interested 😃
Tomorrow, we will present "Open Graph Benchmark: Datasets for ML on graphs" at Session 7 of NeurIPS2020! Come chat with us at the poster!
Spotlight (7:30pm--7:40pm PT):
Poster (9pm--11pm PT):
Website:
Excited to release our paper on the OGB Large-Scale Challenge (OGB-LSC) @ KDD Cup 2021 (on-going)!
Learn about our large datasets + extensive baseline analyses, showing that big expressive models are promising for large-scale graph data!
In OGB-LSC, we have prepared a set of three realistic and *extremely* large-scale graph datasets.
Looking forward to the innovative solutions from the community.
Organized with the amazing team:
@rusty1s
@ren_hongyu
@NakataMaho
@ericdongyx
@jure
We excited to announce OGB-LSC at KDD Cup 2021: A Large-Scale Challenge for Machine Learning on Graphs. Competition ends until June 8. Looking forward to your participation!
@kdd_news
Today Facebook AI and Carnegie Mellon are announcing the Open Catalyst Project, using AI to accelerate quantum mechanical simulations — and thus discover scalable ways to store and use renewable energy.
Excited to share that our Open Graph Benchmark Large-Scale Challenge (OGB-LSC) paper has been accepted to NeurIPS! We have updated our datasets and released public leaderboards:
Paper:
Really excited to announce the winners of OGB-LSC!!
All the winning solutions (code and technical reports) are made publicly available. Learn about the state-of-the-art ML methods for large-scale graph data!
Winners of the OGB-LSC graph machine learning challenge have just been announced at
Congratulations to the winning teams from
@BaiduResearch
,
@DeepMind
,
@Synerise
, Harbin and Dalian Institutes of Technology and USTC!
Excited to announce TGB that expands the scope of OGB into temporal graphs! We welcome the community to explore the fruitful direction with our new benchmark!
I am delighted to announce the Temporal Graph Benchmark (TGB), a collection of challenging and diverse benchmark datasets for temporal graph learning, developed in collaboration with the OGB team!
It's great that DRO community is starting to embrace that naive DRO does not work and structural assumptions are really necessary.
It took us 6 conference rejects until the our negative results paper got accepted...
Almost done with my TA jobs for CS224W ()!
One big highlight was to see students innovating over Open Graph Benchmark datasets (). Impressed to see some SoTA performances achieved in one month!!
The YouTube video of my talk on OGB-LSC is out!
Learn about our large-scale graph datasets, the recent KDD Cup, and the lessons we learned.
Thank you
@myamada0
for the invitation!
I will be giving a talk on the recent OGB Large-Scale Challenge at the KDD Cup 2021. Registration is here:
September 2nd 10am JST
September 1st 6pm PDT
日本の方もぜひ!
Congrats on this high recognition and honored to be part of the GraphCast team during my internship there.
It’s such an important and right problem to apply machine learning to have a real-world impact.
🏅 The winner of the 2024
#MacRobertAward
is...
@GoogleDeepMind
Huge congratulations to the team behind GraphCast, which uses machine learning and vast data sets to give highly accurate and timely weather predictions up to 10 days in advance:
#RAEngAwards
Excited about the release of the Hybrid GNN blog post, from my time at Kumo!
A single, sophisticated GNN model to achieve SoTA on recommendation. Deployed to many customers to generate huge (e.g., $100 million) additional sales. Check it out!