Research explainers on graph neural networks — GCNs, GATs, message passing, and graph-based representation learning.
Picture a university email network where a research group emails constantly among themselves and rarely emails anyone outside the lab. Somewhere in that graph sits a single email, sent maybe once a semester, that connects the lab to a different…
Ricci Flow and Graph Curvature for Advanced Community Detection Read More »
Every brain scan analysis method has to make an early decision that quietly shapes everything downstream. How many regions should the brain be cut into, and how much of the time series should be treated as one chunk. Most systems…
Imagine a fraud detection system built on top of a bank’s transaction graph, where accounts, merchants and devices are all different kinds of nodes connected by different kinds of edges. Now imagine an attacker who does not need to touch…
Backdoor Attacks Now Target Heterogeneous Graph Neural Networks Read More »
Knowledge graph question answering systems typically ask a question, find the relevant topic entity in the graph, expand outward through entity-to-entity relations, and then hand the resulting subgraph to a reasoning module that identifies the answer. That pipeline works well…
A music app might have a handful of global hits that everyone streams and millions of tracks that almost nobody ever plays. A shopping site might have a few bestsellers and a much longer shelf of items with one or…
CHAKG Lets Popular Items Rescue Long Tail Recommendations Read More »
Picture a grocery chain planner staring at three separate spreadsheets on a Monday morning. One says total pasta sales will rise next week. Another, built for the brand level, disagrees. A third, tracking individual items, disagrees with both. None of…
Graph Neural Networks Bring Coherent Forecasts to Retail Read More »
AngioGraphCAD models coronary arteries as graphs, letting a graph neural network assess lesion-level risk the way cardiologists reason — one lesion at a time.
Orthogonal Bases for Equivariant Graph Learning with Provable k-WL Expressive Power | AI Trend Blend Jia He and Maggie X. Cheng from Illinois Institute of Technology have found a way to build GNNs with the same k-WL and k-FWL expressive…
Orthogonal Bases for Equivariant Graph Learning with Provable k-WL Expressive Power Read More »
Improving Graph Neural Networks on Multi-node Tasks with the Labeling Trick | AI Trend Blend A team from Peking University and Georgia Tech has built a formal theory explaining why the most widely used GNN approach to multi-node tasks breaks…
Improving Graph Neural Networks on Multi-node Tasks with the Labeling Trick Read More »
PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks | AI Trend Blend A multi-institution team spanning Peking University, UC San Diego, Stony Brook University, and Weill Cornell Medicine built PEGN — a framework that injects localized…
PEGN: How Persistent Homology Breaks the WL Barrier in Graph Neural Networks Read More »