user avatar
Cristian Bodnar
@crisbodnar
Founding Member @ Project Prometheus, Co-founder @SilurianAI.
Joined May 2014
Posts
  • user avatar
    My PhD Thesis is finally online. I hope it will provide some new perspectives for anyone interested in geometry, topology, deep learning and graph neural networks. repository.cam.ac.uk/items/06b0b8e5…
    🚀Today, I successfully defended my PhD thesis on Topological Deep Learning @Cambridge_CL📖. I was honoured to discuss my work with my examiners, @wellingmax and @jmhernandez233. Very grateful to everyone who has been part of this journey and in particular to @pl219_Cambridge!
  • user avatar
    Replying to @Tim_Denning
    Unknown? The guy literally won a Nobel prize and is behind some of the most influential works in philosophy.
  • user avatar
    🚀Today, I successfully defended my PhD thesis on Topological Deep Learning @Cambridge_CL📖. I was honoured to discuss my work with my examiners, @wellingmax and @jmhernandez233. Very grateful to everyone who has been part of this journey and in particular to @pl219_Cambridge!
  • user avatar
    I’m super excited to share that I have joined the @MSFTResearch AI4Science team as a Senior Researcher to work on deep learning for (bio)chemistry and PDEs 🙌🔬. Can’t wait to work with awesome people and tackle hard problems in these areas! 🚀
  • user avatar
    I remember Geoff Hinton at AAAI 2020 saying how his advisor used to say “Reading rots the mind”. In research, one should simply spend more time thinking how to solve a problem rather than how others solve it. It is one of the best advice I’ve ever received.
  • user avatar
    Neural Sheaf Diffusion (w/ @Francesco_dgv, @b_p_chamberlain, @pl219_Cambridge, @mmbronstein) was accepted to #NeurIPS2022 🎉. Glad to finally attend a conference in-person right at the end of my PhD. The final version & code coming soon (if we survive the ICLR deadline).
    We wrote a blog post providing a friendly overview of our latest work on g̶r̶a̶p̶h̶ sheaf neural networks, heterophily, oversmoothing and expressive power 👇
    GIF
  • user avatar
    Do you want to use geometric DL to solve cutting-edge scientific problems? Then join our PDE team @MSFTResearch AI4Science for an amazing internship opportunity! You will work closely with @jo_brandstetter, @wellingmax and me on large-scale PDE modelling. careers.microsoft.com/us/en/job/1552…
  • user avatar
    So this is what we’ve been up to lately. Extremely excited to build the next generations of frontier models for Earth simulations. Our GFT model is the first step in that direction. More exciting updates coming soon…
    YC S24's @SilurianAI is building foundation models to simulate Earth, starting with weather. From assessing the risk of wildfires to predicting energy grid load, they provide an infrastructure layer for our planet. ycombinator.com/launches/Lcz-s… Congrats on the launch, @rejuvyesh,
    00:00
  • user avatar
    📢Using sheaf neural nets just got a lot easier. We propose pre-processed connection Laplacians as a geometric "diffusion" operator for graphs. This work by @fedzbar w/ Haitz Sáez, @mmbronstein, @PetarV_93 and @pl219_Cambridge will appear at the TAG in ML Workshop at #ICML2022.
  • user avatar
    We wrote a blog post providing a friendly overview of our latest work on g̶r̶a̶p̶h̶ sheaf neural networks, heterophily, oversmoothing and expressive power 👇
    GIF
    In a new blog post authored with @crisbodnar @Francesco_dgv and input from @currying we describe a new approach for studying and designing GNNs using diffusion on cellular sheaves, an exotic object from the field of algebraic topology towardsdatascience.com/neural-sheaf-d…
  • user avatar
    One of the reasons I quit RL a few years ago is that almost nothing creative was happening. Most of the research was (and still is) simply transferring recent successes in representation learning in an RL setting, and that seemed boring to me.
    “What happened to Reinforcement Learning research and labs?” Discussion: old.reddit.com/r/MachineLearn…
  • user avatar
    ML YouTuber @gordic_aleksa made a video explaining our Neural Sheaf Diffusion paper starting from the very basics of sheaf theory. He obviously put a lot of effort into understanding and explaining some very abstract math intuitively, and he did a fantastic job!
    🧠What happens when you combine ideas from algebraic topology (sheaf theory) with Graph Neural Nets? You get Neural Sheaf Diffusion GNNs - provably more expressive! YT: youtube.com/watch?v=JiQmkh… @crisbodnar @Francesco_dgv @b_p_chamberlain @pl219_Cambridge @mmbronstein
  • user avatar
    Want to hear more about how topology can improve GNNs? Tomorrow, 7 Dec 4:30 - 6 PM GMT @ffabffrasca and I will be presenting our #NeurIPS2021 poster on CW Networks. You can also find us (in person) at the Cambridge NeurIPS meetup on Wednesday. Hope to see you at one of these!
  • user avatar
    I’ve recommended this course in all the talks I gave in the past year. IMO, it is incredibly useful for anyone doing Graph ML / Geometric Deep Learning these days (and not only).
    The @CarnegieMellon course on Discrete Differential Geometry is back for Spring 2022! If you want to follow along, all lectures, notes, and coding exercises are online for free at geometry.cs.cmu.edu/ddg The welcome video here gives a quick overview: youtube.com/watch?v=mas-PU…