user avatar
Patrick Kidger
@PatrickKidger
🚀 Doing something new. Curing uncured diseases. 🧪 antibodies+AI, neural ODEs, open-source JAX ecosystem. 👨‍💻 ex-{Cradle, Google, Oxford}.
Mostly Zürich. I travel a lot.
kidger.site
Joined July 2020
241
Following
11K
Followers
Posts
  • Pinned
    user avatar
    Patrick Kidger
    @PatrickKidger
    Feb 8, 2022
    ⚡️ My PhD thesis is on arXiv! ⚡️ To quote my examiners it is "the textbook of neural differential equations" - across ordinary/controlled/stochastic diffeqs. w/ unpublished material: - generalised adjoint methods - symbolic regression - + more! arxiv.org/abs/2202.02435 v🧵 1/n
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Sep 11, 2024
    Had a paper rejected for the 'level of difficulty of the mathematics', and that we should go to 'a less prestigous journal'... 😄 Sorry, we'll be sure to unnecessarily overcomplicate our proofs next time! 🎉
    162K
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Apr 7, 2021
    So I've made a library - "torchtyping" - for annotating a PyTorch Tensor's shape (dtype, names, layout, ...) And at runtime it checks that the annotations are consistent and correct! github.com/patrick-kidger… Bye-bye bugs! Say hello to enforced, clear documentation of your code.
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Jul 4, 2022
    Aaaaannouncing: my first open-source library since I joined Google! 🔥sympy2jax🔥 Build your physics-informed model in SymPy, perform arbitrary symbolic manipulations on it, then convert it to JAX and train it via gradient descent!
    GitHub - patrick-kidger/sympy2jax: Turn SymPy expressions into trainable JAX expressions.
    From github.com
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Jun 21, 2022
    Huge life update: I have finished my PhD at Oxford, and I am joining Google X!🌟(@TheTeamAtX) (Using REDACTED to tackle impossible problems like REDACTED and REDACTED.🕵️‍♂️) Currently at Google London, but I'm the only X-er here. So DM me if you're local, I'd love to make friends!
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Aug 25, 2022
    🌟🌟My team at Google X is hiring! We're looking for someone in {Proteins}∩{ML}, to apply Transformers/LLMs to protein design. (We're doing some really really cool stuff.🤖) DM me + apply here: x.company/careers-at-x/5… This could be you!
    Patrick smiling in the sun, wearing a lot of X stash.
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Feb 20, 2022
    As a bit of an experiment, I've made my next academic poster in HTML+CSS -- not LaTeX! IMO the result looks very pleasant. I much prefer working with CSS over LaTeX when trying to make things look pretty. WDYT?
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Oct 31, 2023
    Use your thesis title as the prompt for #DALLE3. Here goes: "On Neural Differential Equations" It's beautiful!
    user avatar
    Sergey Ovchinnikov
    @sokrypton
    Oct 31, 2023
    Use your PhD thesis title as the prompt 🤓 Is it time to restart this trend? But now with #DALLE3? (Here is mine: "Protein structure determination using evolutionary information")
    158K
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Nov 9, 2023
    🌟 New blog post: Torch-to-JAX! I made the jump PyTorch->JAX a couple of years ago. (JAX is just awesome for the kind of sciML I do!) So for those of you making the same jump, I wrote a JAX quickstart guide for a PyTorch developer. :) Link: kidger.site/thoughts/torch… ⚡️
    kidger.site
    Patrick Kidger
    Personal Website. Math, SciML, scuba diving!
    58K
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Aug 1, 2022
    "jaxtyping" is an amazing way to catch shape errors in your #JAX code. Just add type annotations!🔍 (GitHub: github.com/google/jaxtypi…) And as of the new version, it now supports the holy grail of shape-checking: symbolic expressions!🎉🎉 For example:
  • user avatar
    Patrick Kidger
    @PatrickKidger
    May 12, 2021
    New paper: Neural Rough Differential Equations ! Greatly increase performance on long time series, by using the mathematics of rough path theory. arxiv.org/abs/2009.08295 github.com/jambo6/neuralR… Accepted at #ICML2021! 🧵: 1/n (including a lot about what makes RNNs work)
  • user avatar
    Patrick Kidger
    @PatrickKidger
    May 15, 2022
    💐Happy to announce that our #ICML2022 paper... ... ...nah just kidding it was rejected lol.
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Nov 12, 2024
    For those chasing down this rabbithole, I wrote a 'learn JAX as a PyTorch developer' guide a while ago: kidger.site/thoughts/torch… If that's helpful :)
    user avatar
    xjdr
    Noumena
    @_xjdr
    Nov 10, 2024
    This has been and will continue to be my recommendation for anyone in this position. Learn jax and sign up for sites.research.google/trc/about/ Its one of the best things Google has ever done. You can do meaningful research for free, but the learning curve is steep. strap in
    kidger.site
    Patrick Kidger
    Personal Website. Math, SciML, scuba diving!
    46K
  • user avatar
    Patrick Kidger
    @PatrickKidger
    Feb 14, 2022
    ⭐️Announcing Diffrax!⭐️ Numerical differential equation solvers in #JAX. Very efficient, and with oodles of fun features! GitHub: github.com/patrick-kidger… Docs: docs.kidger.site/diffrax Install: `pip install diffrax` 🧵 1/n