LiMNet: efficient real-time AI inference for high-frequency temporal interactions
Summary
Lightweight Memory Network for processing high-frequency temporal interactions and capturing event causality patterns. Highly scalable: processes 1000s of interactions using single CPU core with less than 1MB memory. Applications include botnet detection, fraud/laundering detection, disease propagation tracking, fake news detection, and recommender systems.
Themes
Keywords
temporal interactions, lightweight ML, botnet detection, fraud detection
Poster
Click image to open full size