This code shows how to train a model in Amazon SageMaker using a custom loss function for a binary classification problem in which the costs of different kinds of misclassification are very different. - View it on GitHub