Big thanks to Jongwoo Kim for his hard work, Woosung Kang, and my advisor, Noseong Park, for his guidance!
On a personal note, all glory to God for the wisdom and faith given throughout this research.
๐ฏ๐๐ ๐๐ ๐๐๐๐๐: We propose ๐ฬฒฬณฬฒ๐จฬฒฬณฬฒ๐ฬฒฬณฬฒ๐ฬฒฬณฬฒ๐ฬฒฬณฬฒ๐ ฬฒฬณฬฒ๐ฬฒฬณฬฒ, which pre-trains on synthetic graphs generated from random graph models in network science with controlled structural properties. One checkpoint, in-context learning at inference. No ๐๐-๐ก๐๐๐๐๐๐๐ or ๐๐๐๐-๐ก๐ข๐๐๐๐ needed for new graphs.
๐พ๐๐๐ ๐๐ ๐๐๐๐๐: GNNs require separate training for each graph due to varying homophily levels, community structures, and feature distributions. Collecting diverse real-world graph data for pre-training is difficult, and using AI-generated synthetic data can be unstable.
Our paper "๐๐๐๐ซ๐ง๐ข๐ง๐ ๐๐จ๐ฌ๐ญ๐๐ซ๐ข๐จ๐ซ ๐๐ซ๐๐๐ข๐๐ญ๐ข๐ฏ๐ ๐๐ข๐ฌ๐ญ๐ซ๐ข๐๐ฎ๐ญ๐ข๐จ๐ง๐ฌ ๐๐จ๐ซ ๐๐จ๐๐ ๐๐ฅ๐๐ฌ๐ฌ๐ข๐๐ข๐๐๐ญ๐ข๐จ๐ง ๐๐ซ๐จ๐ฆ ๐๐ฒ๐ง๐ญ๐ก๐๐ญ๐ข๐ ๐๐ซ๐๐ฉ๐ก ๐๐ซ๐ข๐จ๐ซ๐ฌ" has been accepted to #ICLR2026. See you in Rio! ๐ง๐ท
๐ OpenReview: openreview.net/forum?id=Fmx...
Iโve been recognized as an Outstanding Reviewer (top 10% of reviewers) for both the August and February cycles of #KDD2025 in the research track!
kdd2025.kdd.org/research-tra...
Got the #IJCAI2025 decisions today. Our paper "Learning Advanced Self-Attention for Linear Transformers in the Singular Value Domain" made it in!
The other one got rejected though. Win some, lose some I guess.
The deadlines for #CIKM2025 are coming up in three weeks! There is still time to prepare your papers for submission. Check out all the CFPs and full timeline at cikm2025.org!
Had a presentation on our "Graph Convolutions Enrich the Self-Attention in Transformers!" paper at #NeurIPS2024! Really enjoyed receiving lots of interesting questions about the potential applications of our work.
I totally agree with Sepp Hochreiter's message that "AI does not End with Scaling". I hope my research direction will follow a similar direction. Had great conversations with him after his inspiring invited talk. Grateful for these meaningful interactions at #NeurIPS2024.
Iโll be presenting our work โGraph Convolutions Enrich the Self-Attention in Transformers!โ in #NeurIPS2024 ๐จ๐ฆ!
๐ Location: East Exhibit Hall A-C #2100
๐๏ธ Date: Wed 11 Dec
โฐ Time: ๐ 4:30 p.m. - ๐ข 7:30 p.m. PST.
More details can be found on this #NeurIPS page: lnkd.in/g47GSuZp
I'm making a list of AI for Science researchers on bluesky โ let me know if I missed you / if you'd like to join!
go.bsky.app/AcP9Lix
I made a #starterpack for computational math ๐ป๐งฎ so please
1. share
2. let me know if you want to be on the list!
(I have many new followers which I do not know well yet, so I'm sorry if you follow me and are not on here, but want to - drop me a note and I'll add you!)
go.bsky.app/DXdZkzV
Yeah here's my geometry processing starter pack. I hand-picked researchers and practitioners either in geometry processing or not, and some people are definitely missing so,
go.bsky.app/Dqw66qn
Here is an initial starter pack list on Machine Learning on Graphs: go.bsky.app/HN2MTzp
