The paper “Prescriptive SVD-Inspired Attention via Spectral Energy Retention” has been published in Transactions on Machine Learning Research (TMLR), 2026.
Building on SVD-inspired attention, the work moves from diagnostic interpretation to operational intervention. It proposes a diagnosis-intervention-verification framework and evaluates spectral energy retention in the attention-score pathway, reducing score directions and computational cost while preserving accuracy across several vision benchmarks.
- Authors: Vasileios Arampatzakis, Vasileios Sevetlidis, George Pavlidis
- Journal: Transactions on Machine Learning Research (TMLR), 2026
- arXiv submission date: 21 September 2026
- arXiv: arXiv:2609.24370
You may also like
-
Towards Heritage World Models: from digital twins to predictive heritage systems
-
ICASSP 2026: representation-diverse self-supervision for bioacoustic learning
-
AISTATS 2026: process-tensor tomography of SGD and training memory
-
ICPRAM 2026: two-stage angular alignment for positive-unlabeled learning
-
IEEE Access 2026: interpretable Vision Transformers via SVDA