Excited to share our latest paper 'Unsupervised deep representation learning enables phenotype discovery for genetic association studies of brain imaging' nature.com/articles/s4200… under mentorship of Degui Zhi Dr Degui Zhi, Center for AI and Genome Informatics, McWilliams School of Biomedical Informatics

🥳🥳🥳 The AI for Science slack channel has almost reached 1,000 users, super excited about this growth! Welcome anyone interested in AI for Science to join, chat and post any related events, resources, or hiring info! AI for Science

join.slack.com/t/aiforscience…

This is a super interesting paper, proving once more how important residual connections are. In this case residuals are interpreted as small transformations over time, strongly improving the UNet idea for spatio-temporal problems! I learned a lot when reading this :)

Check out our #ICLR2024 paper, “SineNet: Learning Temporal Dynamics in Time-Dependent Partial Differential Equations” (openreview.net/forum?id=LSYhE…).
U-Nets are widely used for static image segmentation tasks in computer vision. [1/N]

Check out our #ICLR2024 paper, “SineNet: Learning Temporal Dynamics in Time-Dependent Partial Differential Equations” (openreview.net/forum?id=LSYhE…), where we improve the U-Net architecture for modeling time-evolving PDEs. (1/4)

Shout out to the co-first author Jacob Helwig, my advisor Shuiwang Ji and all the collaborators. [8/N, N=8]

(4/4) We open-source our code as part of the AIRS library (github.com/divelab/AIRS/).

Authors: Xuan Zhang, Yuchao Lin, Yaochen Xie, Stephan Wojtowytsch, Cong Fu, Shuiwang Ji

Thrilled to announce that our paper Active Test-Time Adaptation has been accepted by ICLR 2024! 🎉 Many thanks to my great advisor Shuiwang Ji and the remarkable co-author Xiner! Interested in this new TTA direction? Follow our project: 😄
github.com/divelab/ATTA
#ICLR2024

This is the perfect justification for closed source AI. We will see many more “As part of our commitment to releasing our research safely and responsibly, we will not be sharing model weights.” from now on.

Glad to share some new slides based on the Bishop book (bishopbook.com)
Attention, Transformers, and LLMs: people.tamu.edu/~sji/classes/A…
GNN: people.tamu.edu/~sji/classes/G…
Feel free to use and welcome any feedback.

Empowering GNNs via Edge-Aware Weisfeiler-Leman Algorithm

Meng Liu, Haiyang Yu, Shuiwang Ji.

Action editor: Yaoliang Yu.

openreview.net/forum?id=VDy6L…

#gnn #gnn s #graph

Excited to share that ComFormer has been accepted ICLR 2024!

A frame representation based on the lattice of crystals is proposed, which is natural, efficient with O(NK), and geometrically complete with verifications.

One of my favorites along the journey!

TrustLLM: Trustworthiness in Large Language Models

proj: trustllmbenchmark.github.io/TrustLLM-Websi…
repo: github.com/HowieHwong/Tru…
abs: arxiv.org/abs/2401.05561

Here is a list of inexcusable mistakes in slide design:
- not using all the space: showing a diagram or figure that does not fill as much space as possible on your slide. It's very frustrating for the audience to have to squint when there is ample unused space around a figure.
-

Delighted to share InterMat project:
- Large DFT dataset of surfaces and interfaces
- Benchmarked workflow against experimental data
- Unified AI model for electronic band edges
- 1.4 trillion interfaces analyzed with AI
#jarvisnist #semiconductors
arxiv.org/abs/2401.02021

📢 ICYMI, here is the video of the keynote talk I gave last Friday at ASU School of Computing and Augmented Intelligence's #AI day titled 'Can LLMs Really Reason & Plan?'. It is the most comprehensive of my talks on the topic to-date. Slides are available with the video.

youtube.com/watch?v=uTXXYi…

Highlights of Shuiwang Ji’s talk: generative models (g-spherenet, graphbp, symat) are effective for target based drug design, Crystalline Materials, and proteins! Shuiwang Ji GenBio Workshop @Neurips2023

Hope to see you all tomorrow! #NeurIPS2023

🌟 Looking forward to the workshop tomorrow! 🌟

📍 Location: Hall C2
⏰ Starting at 8:15AM

🎙️ 7 speakers + 🤝 panel with funding agencies + 🔬 Open Catalyst Project challenge + 🗣️10 oral talks + 🖼️140(!!) posters

neurips.cc/virtual/2023/w…

Most of NeurIPS paper awards are given to LLMs. Do we still need other ML areas or LLM is all you need?