Scalable large margin pairwise learning algorithms
Classification is a major task in machine learning and data mining applications. Many of these applications involve building a classification model using a large volume of imbalanced data. In such an imbalanced learning scenario, the area under the ROC curve (AUC) has proven to be a reliable performance measure to evaluate a classifier. Therefore, it is desirable to develop scalable learning algorithms that maximize the AUC metric directly. The kernelized AUC maximization machines have established a superior generalization ability compared to linear AUC machines. However, the computational cost ...
(For more, see "View full record.")