I have created a custom loss function to deal with an issue of class imbalance (my datasets have only 2 classes). In my implementation, of a simple neural network with a cus