Briefly training ten MIL models, merging the top three validation performers, and then fully training the merged model reduces run-to-run test AUC variability across five MIL methods on two pathology datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Reducing Variability of Multiple Instance Learning Methods for Digital Pathology
Briefly training ten MIL models, merging the top three validation performers, and then fully training the merged model reduces run-to-run test AUC variability across five MIL methods on two pathology datasets.