Cost Weighted Logistic Loss
The problem of weighting the type 1,2 errors on binary classification came up in a forum I visit. My solution: Here we can see the different loss behaviours We will try to classify the Virginica species from the iris dataset And now the loss functions:
