Proud of my student, Hussien Abdallah, for his dedicated work on his PhD. Our paper "Towards a GML-Enabled #Knowledge #Graph Platform" was accepted to #ICDE2023. We propose KGNet, a #graph #ML (GML) enabled RDF engine with GML as a service to automate the #AI pipelines on KG. KGNet automatically trains GNN models based on a given memory or time budget and maintains a #KG based on metadata collected during the training to help the KGNet query optimizer opt for the near-optimal execution plan of GNN-enabled queries on #KGs based on constraints related to model accuracy and inference time.

Hussien Abdallah is an outstanding candidate if you are looking for a research intern with a great experience in #KG, #GNN, sampling techniques, and exceptional programming skills.

Geometric Deep Learning Grids, Groups, Graphs, Geodesics, and Gauges

https://arxiv.org/abs/2104.13478

#deeplearning #geometriclearning #graphs #graphlearning #gnn

RT @marinkazitnik@twitter.com

Today we share a paper introducing graph neural network #GNN powered mutual interactors to predict #molecular #phenotypes

Excited about this paper because it shows how thoughtful probing of AI model behavior can pave the way to improvements on many downstream tasks [1/6]

🐦🔗: https://twitter.com/marinkazitnik/status/1610328340212207617

ROLAND: A new framework to repurpose static GNNs to dynamic GNNs.

In real life, interaction graphs are often dynamic: edges and nodes change with time 🕸️

ROLAND ia built with PyG @PyTorch GraphGym to efficiently explore the GNN design space.

https://arxiv.org/abs/2208.07239

https://github.com/snap-stanford/roland

#graph #graphneuralnetworks #gnn #neuralnetwork #ann #nn #machinelearning #datascience

ROLAND: A new framework to repurpose static GNNs to dynamic GNNs.

In real life, interaction graphs are often dynamic: edges and nodes change with time 🕸️

ROLAND ia built with PyG @PyTorch GraphGym to efficiently explore the GNN design space.

https://arxiv.org/abs/2208.07239

https://github.com/snap-stanford/roland

#graph #graphneuralnetworks #gnn #neuralnetwork #ann #nn #machinelearning #datascience

Learned about Petar Veličković's work and cats.for.ai yesterday. #categorytheory #GNN #AI #catsforai #machineLearningStreetTalkMlst
This Christmas gift is too cool, it's going to steal spare time.

Day trip to #DESY to check out if we can refurbish an old unused calorimeter for the brandnew #LUXE-NPOD experiment. 4.5h of uninterrupted work time to finally read two recent master theses on #GNN based reconstruction and #AnomalyDetection.

Most #DeepLearning is done on static graph structures.

In real life, interaction graphs are often dynamic: edges and nodes change with time 🕸️

This is where dynamic graph neural networks (#GNN) come in.

ROLAND is a new framework that repurposes static GNNs to dynamic GNNs.

I'm currently attending an Intro to #DeepLearning seminar and the topic of our next session is #GNN. While searching for good intros to graph neural networks I found this ML Tech Talk by Petar Veličković really great https://youtu.be/8owQBFAHw7E His twitter thread about other GNN resources is also gold. https://twitter.com/PetarV_93/status/1306689702020382720

RT @james_y_zou@twitter.com

Very excited to share our new @natBME@twitter.com paper: graph NN discovers #tumor #spatial motifs that predict patient response to cancer treatments

We trained our #GNN on spatial proteomics of 10^6 cells from tumor samples to find disease motifs/subgraphs. Paper https://rdcu.be/cZpgJ
🧵

🐦🔗: https://twitter.com/james_y_zou/status/1590786474404098048

RT @goodnewsnetwork
A Town Near Rome is Selling Old Homes for $1, a Trend Across Italy in Many Beautiful Villages

#GoodNews #GNN #Italy

https://www.goodnewsnetwork.org/1-euro-houses-available-across-italy-with-a-5000-euro-deposit/