“Two-Stage Angular Alignment for Positive-Unlabeled Learning” proposes a geometric weak-supervision method for settings where only positive and unlabeled examples are available.
The work uses angular prototypes and a two-stage curriculum to first compact positives and then repel overly similar unlabeled samples, producing an interpretable representation-based PU learning framework.
- Authors: Vasileios Sevetlidis, George Pavlidis, Antonios Gasteratos
- Venue: 15th International Conference on Pattern Recognition Applications and Methods (ICPRAM 2026), Marbella, Spain
- Conference dates: 2-4 March 2026
- DOI: 10.5220/0013964100004067
- Open record: Zenodo record
You may also like
-
TMLR 2026: prescriptive SVD-inspired attention via spectral energy retention
-
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
-
IEEE Access 2026: interpretable Vision Transformers via SVDA