RT @SebastienBubeck: Enjoy!

@kfountou This classifier is implementable using a message-passing GNN and is the best of both worlds (an MLP for a noisy graph and a GCN for an informative graph) across the range of SNR in the edges/features on synthetic data. Work is pending to make it scalable for use on real data.

@kfountou Following these analyses, we define a precise notion of Bayes optimality for node classification problems and compute the optimal classifier for arbitrary distributions of the node features and edge connectivity.

@kfountou We analyze GNNs from this statistical perspective. We isolate the convolutions from the layers for GCN architectures to understand its variance reduction effects on the data. For GAT, we identify regimes of the SNR of the node features where attention helps or does not help.

RT @kfountou: Positional Attention: Out-of-Distribution Generalization and Expressivity for Neural Algorithmic Reasoning We propose calcul…

RT @kfountou: This paper was just accepted at NeurIPS. I am particularly happy about this because it originated as a course project by Robe…

RT @kfountou: If you are at #ICML2024 and interested in the theory of graph neural networks, come by our poster 'Graph Attention Retrospect…

RT @kfountou: I guess that as of today I can also announce that I have been promoted to the rank of Associate Professor. I am mostly maki…

RT @backdeluca: For those participating in the Complex Networks in Banking and Finance Workshop, I’ll be presenting our work on Local Graph…

@PIBHomeAffairs @narendramodi @AmitShah Thanks for the great initiative! But unfortunately, it doesn't work for me currently. My Indian passport has a Canadian address (enforced by the Indian consulate) and the portal allows only Indian addresses.

RT @kfountou: Paper: Simulation of Graph Algorithms with Looped Transformers (revised version @icmlconf) + Multi-tasking (Remark 6.5) + Di…

RT @kfountou: Paper: Analysis of Corrected Graph Convolutions We study the performance of a vanilla graph convolution from which we remove…

RT @kfountou: .@backdeluca is at ICLR and he will present his joint work with @shenghao_yang on "Local Graph Clustering with Noisy Labels".…

RT @kfountou: .@backdeluca's paper got accepted at ICML. The final version will include a proof for multi-tasking and comments about trai…

RT @WenhuChen: If your model is weak, your paper might end up getting more citations because people are always happy to include your model…

RT @kfountou: Paper: Simulation of Graph Algorithms with Looped Transformers Current empirical results illustrate promising scale generali…

RT @kfountou: Our paper was accepted at ICLR 2024. First paper of @backdeluca as a PhD student in our group!

If you are at #NeurIPS2023 and are interested in 'learning on graphs', please look for our poster in Hall B1+B2 on the 14th (Thursday) from 5-7 pm. Joint work with @kfountou and Aukosh Jagannath.

RT @kfountou: Local Graph Clustering with Noisy Labels, new paper by @backdeluca and @shenghao_yang. Link to the paper: