RT @PietroMonticone: A very nice article in @QuantaMagazine mentions our #EquationalTheories project, led by Terence Tao, which aims to adv….

RT @PietroMonticone: Are you interested in getting started with blueprint-driven formalisation projects in Lean?. I created a template repo….

This project is part of my thesis @CentaiInstitute . Original paper: Original implementation: LinkedIn Post: #GNN #graphneuralnetworks #deeplearning #MachineLearning #PyTorch . 4/4.

Results on the Cora dataset are reproduced, plus a few tweaks were added. Using a modified training strategy on the biased task, the SourceTarget-(V)GAEs surge from around 0.6 AUC to ~0.9 AUC (similarly for AP). 3/4.

Thanks to @gsalhagalvan (@Deezer) for the review and for referencing this implementation in the original one. 2/4.

Check out the Pytorch Geometric (@PyG_Team) implementation of the "Gravity-Inspired Graph Autoencoders for Directed Link Prediction" paper, available here: 1/4.

On the side, thrilled to have engaged with so many #gnn experts during last week's @LogConference meetup in Trento!. Congratulations to the @UniTrento, @FBK_research and @SMLGroup_Trento team for the successful organization!.

Thanks @mljclub and @unito for the opportunity to give this presentation about Graph Deep Learning. We introduced #graphneuralnetworks as an inductive representation learning paradigm on #graphs. Slides:

#JuliaLang #JuliaCon @JuliaConOrg .@JuliaLanguage.

MultilayerGraphs.jl allows for flexible and user-friendly construction of (general) multilayer graphs while retaining compatibility with Graphs.jl and Agents.jl. It has been developed with a "sub-ecosystem" role in mind, enabled by the combination of traits and type hierarchy.

The authors proceed to encode such CAWs using deep learning techniques and prove the superiority of their representation via both inductive and transductive link prediction tasks.

Causal Anonymous Walks may be interpreted as an extension of Anonymous Walks to temporal graphs. Unlike AWs, they retain the dynamic correlations needed to learn the temporal motifs that ultimately drive the network's evolution.

Useful paper from @Abel0828, @jure, @PanLi90769257 et al. on representation learning of temporal graphs via Causal Anonymous Walks: . #GraphML #GraphMining #NetworkScience #MachineLearning.

They prove that their representations effectively retain temporal information via dynamic graph classification tasks, inspired by epidemiological processes.

Their approach allows for lifting static feature map-based graph embedding methods e.g. those derived from RW-based and Weisfeiler–Lehman kernels (maybe many more!) to temporal graphs.

A very interesting paper by Lutz Oettershagen, @nmkriege, @chrsmrrs and Petra Mutzel on static representations of temporal graphs . #GraphML #GraphMining #NetworkScience #MachineLearning.

RT @In_Phy_T: 📢 Here is our talk “Multilayer Network Science in Julia with MultilayerGraphs.jl” presented by @PietroMonticone and @Claudio_….

Last week I completed the #IPSP2023 initiative organized by the Physics Department of @UniTrento , @ConfindustriaTN , @HIT_trentino , @TnSviluppo and others. Amazing people, beautiful location, great work!. @LucaTubiana

👇👇👇.

RT @JuliaConOrg: Thank you to @PietroMonticone and @Claudio__Moroni for your upcoming talk at JuliaCon 2023!. MultilayerGraphs.jl: Multilay….