New VAST method improves semi-supervised learning by decoupling label generation

Researchers have developed VAST (Veracity-Aware Semi-Supervised Training), a novel approach to semi-supervised learning that decouples pseudo-label generation from classifier training. This method infers probabilistic beliefs from the geometry of a frozen self-supervised embedding before distilling them into a classifier, addressing the ill-posed nature of current SSL in cold-start scenarios. VAST utilizes a Veracity Matrix to aggregate label evidence and Veracity Propagation to extend coverage, outperforming existing graph-based SSL baselines across multiple datasets. AI

IMPACT This new method could improve the efficiency and effectiveness of training AI models with limited labeled data.

RANK_REASON The cluster contains a research paper detailing a new method for semi-supervised learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• ScienceCast
• self-supervised embedding
• semi-supervised learning
• VAST
• Veracity-Aware Semi-Supervised Training
• Veracity Matrix
• Veracity Propagation

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →