Delighted to share our review on Graph Machine Learning in Drug Discovery! W/ Relation Therapeutics Thomas Gaudelet Ben Day jyothish soman Jian Tang Michael Bronstein @jptktech and others

arxiv.org/abs/2012.05716

📣 If you’re interested in modelling uncertainty in dynamics, @CristianBodnar, Ben Day, Jacob Moss, Pietro Lio' and I will be discussing our work “Neural ODE Processes” at ICLR poster session 7, Wednesday 9-11 AM BST. 🧵 below: iclr.cc/virtual/2021/p…

Interesting find, Ben Day! I tried other totals like $1.30 and $1.50 and found it still fails, however. Perhaps GPT-3 memorized the original CRT problem?

In this chapter, details of developing a #deeplearning approach to predicting which residues in a protein are involved in forming a PPI—a task known as PPI site prediction—are outlined.
Arian Jamasb Ben Day Cătălina Cangea
Pietro Lio'
#AI #MachineLearning #NLP

Interested in Deep Learning? Comp Bio? Protein-protein interactions? PPI Sites?

Our book chapter Deep Learning for Protein–Protein Interaction Site Prediction is out today! 🥳

Link: link.springer.com/protocol/10.10…

With: Ben Day Cătălina Cangea Pietro Lio' Tom Blundell

Ben Day Arian Jamasb Marta Garnelo In the transductive setup, MPNP achieves top results on 3-class and on all <30% context rates on 7-class; MPNPs are outperformed by labelprop on 11-class (but still beat NPs), reinforcing the findings of the recent work (arxiv.org/abs/2010.13993) by Qian Huang Horace He et al. 10/

Ben Day It’s not identical. Double Moderna with Pfizer boost is better than triple Moderna (the study below is with 3 full doses of Moderna; Moderna booster is half as large and would do worse than shown here):

Stefan Schubert Ben Day 1st doses week to 7-Feb — 2m AZ, 0.9m Pfizer (~70%, 30%}

W/ Pavol Drotar Ben Day Cătălina Cangea Pietro Lio'

Reminder: this is ~2h from now! Join me (virtually) CambridgeCL for a tour-de-GNNs, which I hope will be of use to newcomers as well as seasoned practitioners!

Happy that Ben Day will chair the Q&A! Please reply here w any questions you have, anytime during the talk. (1/3)

Stefan Schubert Ben Day New data from MHRA: implied mix for the most recent week is 56% AZ; 44% Pfizer. AFAIK these releases are only way of tracking the UK mix (updated weekly with 2-week lag) — doesn't seem to be widely known. gov.uk/government/pub…

@sam_power_825 Jack Butcher Ben Day

Ben Day Congrats 🎉🎉🎉

Pierre Ablin Michael E. Sander Mathieu Blondel Gabriel Peyré Nice work! This seems very similar to previous works investigating Second Order Neural ODEs (papers.nips.cc/paper/2020/fil…, proceedings.neurips.cc/paper/2020/fil…), which would be useful to include in the related work.

@CristianBodnar Ben Day Nikola Simidjievski Michael Poli Stefano Massaroli

We love to see it Ben Day ! Thanks for joining us!

Late to the party, but happy to announce our paper 'Neural ODE Processes' was accepted to #ICLR2021 🎉

w/ cristianbodnar Ben Day Jacob Moss Pietro Lio'

openreview.net/forum?id=27acG…

Stay tuned for a full thread

Our work on Neural ODE Processes was accepted to #ICLR2021 . We propose a new type of stochastic process defined over time based on Neural ODEs. Enhanced paper and twitter-thread coming soon.

w/ Alexander Norcliffe Ben Day Jacob Moss Pietro Lio' openreview.net/forum?id=27acG…

Ben Day Stewart Abbott Cambridge University Chris Packham RSPB Doing work in winter or after a thorough check by professionals is better. Leaving nets up until the company gets around to it at some unspecified point is lazy, looks terrible and allows for netting to become loose/weathered and trap wildlife.

Ben Day roon If this film takes place in an alternate universe where there's an evolved linguistic code that can rewire the brain and give it powers the body doesn't physiologically posses, and aliens cracked it, then cool, I just thought it was meant to take place in our current universe.