macro-AUC
A multiclass evaluation metric computed by calculating AUC separately for each class and then taking the arithmetic mean, treating each class equally and being sensitive to the performance of tail classes.
Timeline (last 90 days)
The gates' hard-zero outputs pushed the cascade model's macro-AUC down to 0.747.
Macro-AUC is the macro-averaged version of AUC: AUC is computed separately for each class and then arithmetically averaged across all classes, making it more sensitive to the performance on tail classes.
All Facts (2)
The gates' hard-zero outputs pushed the cascade model's macro-AUC down to 0.747.
50%UnverifiedMacro-AUC is the macro-averaged version of AUC: AUC is computed separately for each class and then arithmetically averaged across all classes, making it more sensitive to the performance on tail classes.
50%