Скачать с ютуб Graph Neural Networks for Link Prediction with Subgraph Sketching в хорошем качестве

Graph Neural Networks for Link Prediction with Subgraph Sketching

Try datamol.io - the open source toolkit that simplifies molecular processing and featurization workflows for machine learning scientists working in drug discovery: https://datamol.io/ Never miss another LoGG Talk, add the schedule to your calendar: https://m2d2.io/talks/m2d2/about/ Join the Learning on Graphs and Geometry Reading Group on Slack: https://join.slack.com/t/logag/shared... Abstract: 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 triangles (the backbone of most LP heuristics) and because they can not distinguish automorphic nodes (those having identical structural roles). Both expressiveness issues can be alleviated by learning link (rather than node) representations and incorporating structural features such as triangle counts. Since explicit link representations are often prohibitively expensive, recent works resorted to subgraph-based methods, which have achieved state-of-the-art performance for LP, but suffer from poor efficiency due to high levels of redundancy between subgraphs. We analyze the components of subgraph GNN (SGNN) methods for link prediction. Based on our analysis, we propose a novel full-graph GNN called ELPH (Efficient Link Prediction with Hashing) that passes subgraph sketches as messages to approximate the key components of SGNNs without explicit subgraph construction. ELPH is provably more expressive than Message Passing GNNs (MPNNs). It outperforms existing SGNN models on many standard LP benchmarks while being orders of magnitude faster. However, it shares the common GNN limitation that it is only efficient when the dataset fits in GPU memory. Accordingly, we develop a highly scalable model, called BUDDY, which uses feature precomputation to circumvent this limitation without sacrificing predictive performance. Our experiments show that BUDDY also outperforms SGNNs on standard LP benchmarks while being highly scalable and faster than ELPH. Speakers: Ben Chamberlain -   / benjaminchamberlain   Sergey Shirobokov -   / sergey-shirobokov   Twitter Hannes:   / hannesstaerk   Twitter Dominique:   / dom_beaini   Twitter datamol.io:   / datamol_io   ~ Chapters 00:00 - Intro 04:03 - Production Model at Twitter 07:32 - First Attempt with GNNs for Recommendations 10:59 - Automorphic Nodes 15:08 - Subgraph GNNs 18:22 - Analyzing Subgraph GNNs: Model Components 22:18 - Subgraph Sketching: Minhash 26:51 - GNNs with Subgraph Sketching 34:10 - BUDDY: Scaling ELPH 37:42 - Results 39:09 - Q+A