New arXiv analysis of 37 graph-based cyber attack detection studies (2019–2026) finds automated representation learning dominates at 73.0%, versus 27.0% handcrafted strategies, highlighting a clear shift toward learned…

#AI #CyberSecurity #GraphML #DataScience
https://arxiv.org/abs/2610.04019

Plain Transformers can rival Graph Neural Networks on link prediction by attending to sampled subgraphs, offering a scalable alternative without heavy structural encodings. arXiv:2602.01553

#AI #GraphML #Transformers
https://arxiv.org/abs/2602.01553