chaitjo

3 ms·

1.
by chaitjo 9mo ago
Author here. I recently defended my PhD thesis on Geometric Deep Learning. This post summarizes 4 years of research into three main questions covering: - Theoretical expressivity of 3D molecular representations (Geometric WL test). - Unifie
2.
An AI researcher in the Cathedral of molecular biology(chaitjo.substack.com)
1 points by chaitjo 10mo ago|1 comments
3.
by chaitjo 10mo ago
AI for science is very hot right now. I wrote some personal reflections about being physically embedded in a world-leading molecular biology lab as an AI researcher: learning to communicate with experimentalists, back-breaking wet lab work,
4.
Hackable AlphaFold 3 without Docker or MSAs(github.com)
3 points by chaitjo 1y ago|1 comments
5.
by chaitjo 1y ago
his is a lightweight, hackable way to run AlphaFold 3 that lets you experiment without the massive MSA databases or Docker overhead. Perfect for tinkering on your laptop or single GPU server - just install, drop in your sequences, and start
6.
Equivariance is dead, long live equivariance?(chaitjo.substack.com)
1 points by chaitjo 1y ago|0 comments
7.
A Hitchhiker's Guide to Geometric GNNs for 3D Atomic Systems(arxiv.org)
5 points by chaitjo 3y ago|1 comments
8.
by chaitjo 3y ago
We’ve just released a “Hitchhiker’s Guide” for getting started with deep neural nets for 3D structural biology & chemistry -- we think it will be useful for newcomers to start their learning journey on the core architectures powering re
9.
by chaitjo 4y ago
Geometric GNNs are an emerging class of GNNs for spatially embedded graphs in scientific and engineering applications, s.a. biomolecular structure, material science, and physical simulations. Notable examples include SchNet, DimeNet, Tensor
10.
A Gentle Introduction to Geometric Graph Neural Networks(github.com)
1 points by chaitjo 4y ago|1 comments
11.
by chaitjo 7y ago
Indeed, the two papers came out within months of each other iirc. The GAT paper discusses Transformers in the context of stabilizing the learning of attention mechanisms. Of course, this connection may be trivial to most people, but I hadn&
12.
by chaitjo 7y ago
In graph terms, Positional encodings are useful for adding sequential/temporal properties to each node in the graph. Indeed, there are works on position-aware GNNs.
13.
by chaitjo 7y ago
Not necessarily. In fact, applying the Transformer/GNN on a full graph is seen by some people as 'discovering' or identifying some useful underlying graph structure. We had an interesting discussion on this on Twitter: https
14.
by chaitjo 7y ago
Thanks for highlighting! (Admin of the lab website)
15.
by chaitjo 7y ago
I am guessing he means that Transformers 'discover' parse trees in sentences, as they operate on fully-connected graphs but end up focusing on the most important connections through attention.