ML at @RelationRx previously @Cambridge_CL, @InfAtEd, @MRC_Epid, Trying to learn.
Joined June 2020
- Tweets43
- Following221
- Followers260
- Likes164
Paul Scherer retweeted
Making AI useful in the lab: the missing math of active learning
A call for greater first-principles theory to power lab-in-the-loop science
with @BlackHC and @paulmorio
Paul Scherer retweeted
🧵1/7 Introducing an “Encyclopedia” of Molecular Design with Machine Learning @NatMachIntell! nature.com/articles/s42256-0…
Collaboration with @arian_jamasb*, @JeffGuo__*, @TianfanFu, @charlieharris01, @yingheng_wang, @chenru_duan, @pl219_Cambridge, @pschwller, and Tom Blundell.
Paul Scherer retweeted
I’m really excited to share our ICLR paper proposing Controlled Monte Carlo Diffusions (CMCD), a new sampler that enforces boundary conditions on the marginals exactly, achieving SOTA scores on ELBO and ln Z estimation!
Thread on neat theoretical contributions and video (1/11)!
It's been a long time since I've had the chance to talk about GRL on dynamic graphs, so it was fun to revisit the topic a little. There's definitely still lots of research and engineering work to do in this area.
Thanks for letting me present and the picture @PetarV_93 !
Now that I've handed in the thesis I've decided to go refactor some of my first projects.
Hahaha now I need a time machine and a crowbar.
Paul Scherer retweeted
Don't miss the next AI in Medicine Seminar to learn about the latest trends and explore collaborations in the exciting field of AI in Medicine.
🗓️ Tues 27 June 11:00am
Seminar Room 12 School of Clinical Medicine @CamBioCampus
👇
Paul Scherer retweeted
Check out our new blog post summarising key advances in Temporal Graph Learning over the last 12 months!
Written with amazing co-authors @shenyangHuang @michael_galkin
towardsdatascience.com/tempo…
My work on "Distributed representations of graphs for drug pair scoring" with Professors @pl219_Cambridge and Mateja Jamnik got accepted as a poster @LogConference!
Come say hi at the virtual poster session today, I will also be at the local meetup in Cambridge tomorrow :)
ArXiv: arxiv.org/abs/2209.09383
OpenReview camera ready: openreview.net/pdf?id=IP-TIS…
I will be at NeurIPS from Monday onwards and would love to meet as many people as possible. If you are around and want to chat about anything, please DM/e-mail/etc.
Paul Scherer retweeted
I’m thrilled to announce that @carbon_re has raised a $4.8m seed round! We’ve been sitting on this news for a while to sort out all the legal bits, but we can finally share it around 🚀
It will help us grow the team, serve more customers, and kickstart an AI research group!
1/🧵
Paul Scherer retweeted
A new technique keeps cells alive during extraction to track the activity of thousands of genes across time. The breakthrough method is called Live-seq and was jointly developed by scientists at @epflSV and @ETH_en actu.epfl.ch/news/live-seq-s…
Paul Scherer retweeted
Announcing the 2022 NeurIPS Workshop on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems!
Please consider submitting a 4-page extended abstract by September 22nd!
gp-seminar-series.github.io/…
Thanks to amazing organizers @gpleiss, @liza_p_semenova, @ziwphd
ALT NeurIPS Workshop on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems
So I have been working on PyRelationAL and it's finally public!
It's a Python active learning library designed to help you research and develop existing and novel(!) AL strategies quickly and reliably.
Don't know AL or where to start? Check the docs!
pyrelational.readthedocs.io/…
The Relation tech team are developing world leading Active Learning capabilities for drug discovery.
To this end, we are
1. Open sourcing internal library PyRelationAL github.com/RelationRx/pyrela…
2. Releasing PoC screens arxiv.org/abs/2202.04202
3. Building a lab
Also, we're hiring!
Of course none of this would have been at all possible without help from @thomasgaudelet @a_ppln @wavybatman @jyo_soman Alice @JPTK_TechBio and everyone else at @RelationRx.
I hope it will be useful to researchers studying this exciting problem and lead to new developments :)
Very nicely written, accessible, and concise survey on GRL for Biomedicine by @_michellemli, @KexinHuang5, and @marinkazitnik . Presents a number of cool avenues for further research if you are early in your PhD or looking to start a new project.
arxiv.org/abs/2104.04883
Paul Scherer retweeted
JAXopt v0.3 has been released! New features: LBFGS, nonlinear conjugate gradient, and Gauss-Newton algorithm! github.com/google/jaxopt
🥳🎉
Congratulations to all the authors for their papers nominated for best resource paper at #CIKM2021
Special congratulations to the best resource paper: "PyTorch Geometric Temporal: Spatiotemporal Signal Processing with Neural Machine Learning Models"
Paul Scherer retweeted
I am super excited to announce PyG 2.0 - a new major release of PyTorch Geometric brought to you by researchers from TU Dortmund and Stanford University.
This new release brings full heterogeneous graph support, GraphGym, and many other features to PyG. bit.ly/3E6cpe9
