@nhammerla

CTO @coding_bio. Ex Cortex @TwitterUK. Ex @babylonhealth #ML #climbing

 London, England
Joined June 2020
  • Tweets62
  • Following280
  • Followers265
1Media
If you are attending #ICLR2023 come to see our “royal flash”: oral on scalable link prediction, spotlight on hyperbolic RL, and poster on gradient gating in GNNs @DrBPChamberlain @tk_rusch @edo_cet @emaros96 @ffabffrasca @thomasmarkovich @hansmire @nhammerla @SergeyShir994
1
15
60
7,954
1/3 Twitter graph ML full house at #ICLR2023 : 📜Poster arxiv.org/abs/2210.00513 🔦Spotlight arxiv.org/abs/2210.01542 top 25% 👄Oral arxiv.org/abs/2209.15486 top 5% spanning my recent interests in continuous GNNs, hyperbolic geometry (applied to RL this time) & link prediction

Gradient Gating for Deep Multi-Rate Learning on Graphs

We present Gradient Gating (G$^2$), a novel framework for improving the performance of Graph Neural Networks (GNNs). Our framework is based on gating the output of GNN layers with a mechanism for...

arxiv.org
5
33
3
165
25,225
Nils Hammerla retweeted
Emanuele Rossi@emaros9623 Jan 2023
Our paper on scaling link prediction to graphs with tens of millions of nodes using subgraph sketching has been awarded an oral presentation (top 5%) at ICLR! arxiv.org/abs/2209.15486 @mmbronstein @DrBPChamberlain @SergeyShir994 @ffabffrasca @hansmire @nhammerla @thomasmarkovich

Graph Neural Networks for Link Prediction with Subgraph Sketching

Many Graph Neural Networks (GNNs) perform poorly compared to simple heuristics on Link Prediction (LP) tasks. This is due to limitations in expressive power such as the inability to count...

arxiv.org
3
28
5
200
24,311
Replying to @emaros96
@emaros96 @ffabffrasca @Francesco_dgv @b_p_chamberlain The best graph ML team attending NeurIPS
3
1
43
Nils Hammerla retweeted
Frank Noe@FrankNoeBerlin27 Nov 2022
Tweeps at #NeurIPS2022 #ML4PS2022 @lrml_bio @AI_for_Science looking for exciting jobs in deep learning and molecular sciences? @MSFTResearch AI4Science is hiring in Berlin and other sites. DM me or approach my colleagues for more info.
37
118
🫡
1
20
Transparency is important to trust. Birdwatch’s algorithm is publicly available in GitHub, along with all of the data that powers it. Whether you are a researcher, a hobbyist, or just curious, you can run Birdwatch’s algorithm on your own computer: twitter.github.io/birdwatch/…
1
9
1
41
1/4 Today we released two new link prediction models developed over the past 12m. We solve many problems that prevent GNNs succeeding at link prediction by adding a message passing mechanism that uses sketches of subgraphs. arxiv.org/abs/2209.15486

Graph Neural Networks for Link Prediction with Subgraph Sketching

Many Graph Neural Networks (GNNs) perform poorly compared to simple heuristics on Link Prediction (LP) tasks. This is due to limitations in expressive power such as the inability to count...

arxiv.org
2
37
3
181
We have two papers accepted at #NeurIPS2022 on neural sheaf diffusion and subgraph GNNs Congratulations to amazing coauthors @Francesco_dgv @crisbodnar @b_p_chamberlain Pietro Lio' @ffabffrasca @HaggaiMaron @beabevi_
10
13
264
Replying to @Airbnb
@Airbnb is using our SIGN architecture to compute graph embeddings at scale! They chose SIGN to build a maintainable and effective pipeline producing features resulting in "significant gains" in their downstream models. More from @AirbnbEng : medium.com/airbnb-engineerin…

Graph Machine Learning at Airbnb

How Airbnb is leveraging graph neural networks to up-level our machine learning

medium.com
2
8
56
Nils Hammerla retweeted
Emanuele Rossi@emaros9621 Jun 2022
Replying to @Airbnb
@Airbnb is using our SIGN method in production to compute graph-based user embeddings, leading to "notable performance gains". Why was it chosen? Scalability, simplicity and predictive performance. medium.com/airbnb-engineerin… @AirbnbEng

Graph Machine Learning at Airbnb

How Airbnb is leveraging graph neural networks to up-level our machine learning

medium.com
3
8
1
33
Here it is: the first Learning on Graphs Conference! 🎊 We think this new venue will be valuable for the Graph/Geometric Machine Learning community. What makes it so important+unique? See our blog post! michael-bronstein.medium.com… 1/6
4
281
37
835
This new partnership will create exciting opportunities for our researchers to collaborate with @twittereng researchers on the development of responsible, community-driven machine learning systems. mila.quebec/en/twitter-annou…

Twitter announces partnership with Mila | Mila

Mila is happy to announce a new partnership with Twitter and the Twitter Cortex data and machine learning (ML) team. This new collaboration will focus on open research engagements between students...

mila.quebec
2
16
2
120
New post @TDataScience about physics-inspired "continuous" graph ML models 🧵 bit.ly/3pxxx7q
4
79
12
349
Nils Hammerla retweeted
Emanuele Rossi@emaros963 Feb 2022
While most GNNs assume a full set of features for all nodes, real-world graphs often have partially missing node features. In a new blog post with @mmbronstein, we discuss how we can learn on such graphs using Feature Propagation (10 min read) 📝 towardsdatascience.com/learn…
5
76
376
1/4 Hope to see friends new & old tomorrow 4:30pm GMT / 8:30am PT poster session 6 #NeurIPS2021 @_JRowbottom @Francesco_dgv and I will occupy the prime virtual real estate known as Spot E3 with BLEND neurips.cc/virtual/2021/post…
3
13
40
Nils Hammerla retweeted
Emanuele Rossi@emaros961 Dec 2021
We introduce Feature Propagation (FP), a new method to learn on graphs with partially missing node features. With 99% of features missing, FP loses only 4% rel. acc. compared to when all features are available. Bonus: It scales to graphs with >100M nodes arxiv.org/abs/2111.12128
8
52
9
307
After a hiatus, a new series of blogs posts. Do differential geometry and algebraic topology sound too exotic for ML? In recent works, we show that tools from these fields bring a new perspective on graph neural networks First post in the series: towardsdatascience.com/graph…
15
195
12
1,038
Nils Hammerla retweeted
Claire@claireewoodcock9 Nov 2021
1/ When do people need an explanation from an AI system? What makes them trust the answer? Excited to share. jmir.org/2021/11/e29386/ #ExplainableAI #XAI #healthtech with @b_mittelstadt @danbusbridge & Grant Blank Spoiler: AI should communicate with humans like they're... human.

The Impact of Explanations on Layperson Trust in Artificial Intelligence–Driven Symptom Checker...

Background: Artificial intelligence (AI)–driven symptom checkers are available to millions of users globally and are advocated as a tool to deliver health care more efficiently. To achieve the...

jmir.org
2
5
1
39
Load more