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

TL;DR

This paper introduces AdaCSL, an adaptive cost-sensitive learning algorithm for neural networks that reduces misclassification costs by adjusting the loss function based on local training-test class distribution mismatches, improving performance on imbalanced and balanced datasets.

Contribution

The paper proposes AdaCSL, a novel adaptive loss function adjustment method for neural networks addressing misclassification costs with theoretical guarantees and empirical validation.

Findings

AdaCSL improves cost metrics on various datasets.

Deep neural networks with AdaCSL outperform other methods.

The approach effectively handles class imbalance and distribution mismatch.

Abstract

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