Adaptive Cost-Sensitive Learning in Neural Networks for Misclassification Cost Problems

We design a new adaptive learning algorithm for misclassification cost\nproblems that attempt to reduce the cost of misclassified instances derived\nfrom the consequences of various errors. Our algorithm (adaptive cost sensitive\nlearning - AdaCSL) adaptively adjusts the loss function such that the\nclassifier bridges the difference between the class distributions between\nsubgroups of samples in the training and test data sets with similar predicted\nprobabilities (i.e., local training-test class distribution mismatch). We\nprovide some theoretical performance guarantees on the proposed algorithm and\npresent empirical evidence that a deep neural network used with the proposed\nAdaCSL algorithm yields better cost results on several binary classification\ndata sets that have class-imbalanced and class-balanced distributions compared\nto other alternative approaches.\n