A spectral and approximate message passing method for label-free aggregation of pre-trained models, robust to heteroskedastic noise and non-informative models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
U-aggregation: Unsupervised Aggregation of Multiple Learning Algorithms
A spectral and approximate message passing method for label-free aggregation of pre-trained models, robust to heteroskedastic noise and non-informative models.