New method constructs diverse decision sources for improved pseudo-label learning

Researchers have developed a method to improve pseudo-label learning by constructing more informative decision sources. Instead of relying on multiple identical models, this approach modifies the internal structure of a shared graph representation to create diverse evidence. By selecting a complementary subset of these constructed sources, the method enhances the precision of pseudo-labels, leading to competitive downstream accuracy on datasets like Cora, CiteSeer, and PubMed. This work highlights the importance of source construction over simply increasing model count for consensus-based pseudo-label learning. AI

IMPACT This research could lead to more efficient and accurate training of machine learning models by improving the quality of pseudo-labels.

RANK_REASON Academic paper detailing a new method for improving pseudo-label learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CiteSeerX
• Connected Papers
• Cora
• DagsHub
• Gotit.pub
• graph convolutional network
• Hugging Face
• IArxiv
• Influence Flower
• Litmaps
• PubMed
• ScienceCast
• scite Smart Citations

• paper
• other

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